sw: add printf

We'll use this to print debugging information out the serial port.

Signed-off-by: Sean Cross <sean@xobs.io>

sw/include/printf.h

@@ -0,0 +1,117 @@
+///////////////////////////////////////////////////////////////////////////////
+// \author (c) Marco Paland (info@paland.com)
+//             2014-2019, PALANDesign Hannover, Germany
+//
+// \license The MIT License (MIT)
+//
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+// 
+// The above copyright notice and this permission notice shall be included in
+// all copies or substantial portions of the Software.
+// 
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+// THE SOFTWARE.
+//
+// \brief Tiny printf, sprintf and snprintf implementation, optimized for speed on
+//        embedded systems with a very limited resources.
+//        Use this instead of bloated standard/newlib printf.
+//        These routines are thread safe and reentrant.
+//
+///////////////////////////////////////////////////////////////////////////////
+
+#ifndef _PRINTF_H_
+#define _PRINTF_H_
+
+#include <stdarg.h>
+#include <stddef.h>
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/**
+ * Output a character to a custom device like UART, used by the printf() function
+ * This function is declared here only. You have to write your custom implementation somewhere
+ * \param character Character to output
+ */
+void _putchar(char character);
+
+
+/**
+ * Tiny printf implementation
+ * You have to implement _putchar if you use printf()
+ * To avoid conflicts with the regular printf() API it is overridden by macro defines
+ * and internal underscore-appended functions like printf_() are used
+ * \param format A string that specifies the format of the output
+ * \return The number of characters that are written into the array, not counting the terminating null character
+ */
+#define printf printf_
+int printf_(const char* format, ...);
+
+
+/**
+ * Tiny sprintf implementation
+ * Due to security reasons (buffer overflow) YOU SHOULD CONSIDER USING (V)SNPRINTF INSTEAD!
+ * \param buffer A pointer to the buffer where to store the formatted string. MUST be big enough to store the output!
+ * \param format A string that specifies the format of the output
+ * \return The number of characters that are WRITTEN into the buffer, not counting the terminating null character
+ */
+#define sprintf sprintf_
+int sprintf_(char* buffer, const char* format, ...);
+
+
+/**
+ * Tiny snprintf/vsnprintf implementation
+ * \param buffer A pointer to the buffer where to store the formatted string
+ * \param count The maximum number of characters to store in the buffer, including a terminating null character
+ * \param format A string that specifies the format of the output
+ * \param va A value identifying a variable arguments list
+ * \return The number of characters that COULD have been written into the buffer, not counting the terminating
+ *         null character. A value equal or larger than count indicates truncation. Only when the returned value
+ *         is non-negative and less than count, the string has been completely written.
+ */
+#define snprintf  snprintf_
+#define vsnprintf vsnprintf_
+int  snprintf_(char* buffer, size_t count, const char* format, ...);
+int vsnprintf_(char* buffer, size_t count, const char* format, va_list va);
+
+
+/**
+ * Tiny vprintf implementation
+ * \param format A string that specifies the format of the output
+ * \param va A value identifying a variable arguments list
+ * \return The number of characters that are WRITTEN into the buffer, not counting the terminating null character
+ */
+#define vprintf vprintf_
+int vprintf_(const char* format, va_list va);
+
+
+/**
+ * printf with output function
+ * You may use this as dynamic alternative to printf() with its fixed _putchar() output
+ * \param out An output function which takes one character and an argument pointer
+ * \param arg An argument pointer for user data passed to output function
+ * \param format A string that specifies the format of the output
+ * \return The number of characters that are sent to the output function, not counting the terminating null character
+ */
+int fctprintf(void (*out)(char character, void* arg), void* arg, const char* format, ...);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif  // _PRINTF_H_

sw/include/usb.h

@@ -7,6 +7,25 @@ extern "C" {
 
 struct usb_setup_request;
 
+enum epfifo_response {
+    EPF_ACK = 0,
+    EPF_NAK = 1,
+    EPF_NONE = 2,
+    EPF_STALL = 3,
+};
+
+// Note that our PIDs are only bits 2 and 3 of the token,
+// since all other bits are effectively redundant at this point.
+enum USB_PID {
+    USB_PID_OUT   = 0,
+    USB_PID_SOF   = 1,
+    USB_PID_IN    = 2,
+    USB_PID_SETUP = 3,
+};
+
+#define USB_EV_ERROR 1
+#define USB_EV_PACKET 2
+
 void usb_isr(void);
 void usb_init(void);
 void usb_connect(void);

sw/src/crt0-vexriscv.S

@@ -6,13 +6,6 @@
 
 _start:
   j crt_init
-  nop
-  nop
-  nop
-  nop
-  nop
-  nop
-  nop
 
 .section .text
 .global  trap_entry
@@ -55,7 +48,6 @@ trap_entry:
   mret
   .text
 
-
 crt_init:
   la sp, _fstack + 4
   la a0, trap_entry
@@ -87,5 +79,6 @@ bss_done:
   csrw mie,a0
 
   call main
-infinit_loop:
+
-  j infinit_loop
+infinite_loop:
+  j infinite_loop

sw/src/div.S

@@ -0,0 +1,121 @@
+  .text
+  .align 2
+
+#ifndef __riscv64
+/* Our RV64 64-bit routines are equivalent to our RV32 32-bit routines.  */
+# define __udivdi3 __udivsi3
+# define __umoddi3 __umodsi3
+# define __divdi3 __divsi3
+# define __moddi3 __modsi3
+#else
+  .globl __udivsi3
+__udivsi3:
+  /* Compute __udivdi3(a0 << 32, a1 << 32); cast result to uint32_t.  */
+  sll    a0, a0, 32
+  sll    a1, a1, 32
+  move   t0, ra
+  jal    __udivdi3
+  sext.w a0, a0
+  jr     t0
+
+  .globl __umodsi3
+__umodsi3:
+  /* Compute __udivdi3((uint32_t)a0, (uint32_t)a1); cast a1 to uint32_t.  */
+  sll    a0, a0, 32
+  sll    a1, a1, 32
+  srl    a0, a0, 32
+  srl    a1, a1, 32
+  move   t0, ra
+  jal    __udivdi3
+  sext.w a0, a1
+  jr     t0
+
+  .globl __modsi3
+  __modsi3 = __moddi3
+
+  .globl __divsi3
+__divsi3:
+  /* Check for special case of INT_MIN/-1. Otherwise, fall into __divdi3.  */
+  li    t0, -1
+  beq   a1, t0, .L20
+#endif
+
+  .globl __divdi3
+__divdi3:
+  bltz  a0, .L10
+  bltz  a1, .L11
+  /* Since the quotient is positive, fall into __udivdi3.  */
+
+  .globl __udivdi3
+__udivdi3:
+  mv    a2, a1
+  mv    a1, a0
+  li    a0, -1
+  beqz  a2, .L5
+  li    a3, 1
+  bgeu  a2, a1, .L2
+.L1:
+  blez  a2, .L2
+  slli  a2, a2, 1
+  slli  a3, a3, 1
+  bgtu  a1, a2, .L1
+.L2:
+  li    a0, 0
+.L3:
+  bltu  a1, a2, .L4
+  sub   a1, a1, a2
+  or    a0, a0, a3
+.L4:
+  srli  a3, a3, 1
+  srli  a2, a2, 1
+  bnez  a3, .L3
+.L5:
+  ret
+
+  .globl __umoddi3
+__umoddi3:
+  /* Call __udivdi3(a0, a1), then return the remainder, which is in a1.  */
+  move  t0, ra
+  jal   __udivdi3
+  move  a0, a1
+  jr    t0
+
+  /* Handle negative arguments to __divdi3.  */
+.L10:
+  neg   a0, a0 
+  bgez  a1, .L12      /* Compute __udivdi3(-a0, a1), then negate the result.  */
+  neg   a1, a1
+  j     __divdi3      /* Compute __udivdi3(-a0, -a1).  */
+.L11:                 /* Compute __udivdi3(a0, -a1), then negate the result.  */
+  neg   a1, a1
+.L12:
+  move  t0, ra
+  jal   __divdi3
+  neg   a0, a0
+  jr    t0
+
+  .globl __moddi3
+__moddi3:
+  move   t0, ra
+  bltz   a1, .L31
+  bltz   a0, .L32
+.L30:
+  jal    __udivdi3    /* The dividend is not negative.  */
+  move   a0, a1
+  jr     t0
+.L31:
+  neg    a1, a1
+  bgez   a0, .L30
+.L32:
+  neg    a0, a0
+  jal    __udivdi3    /* The dividend is hella negative.  */
+  neg    a0, a1
+  jr     t0
+
+#ifdef __riscv64
+  /* continuation of __divsi3 */
+.L20:
+  sll   t0, t0, 31
+  bne   a0, t0, __divdi3
+  ret
+#endif

sw/src/main.c

@@ -1,10 +1,11 @@
-#include <stdio.h>
+#include <fomu/csr.h>
 #include <irq.h>
 #include <usb.h>
 #include <time.h>
 #include <rgb.h>
 #include <spi.h>
-#include <fomu/csr.h>
+#include <tester.h>
+#include <usb-cdc.h>
 
 struct ff_spi *spi;
 

sw/src/mul.S

@@ -0,0 +1,26 @@
+  .text
+  .align 2
+
+#ifdef __riscv64
+#define _RISCV_SZPTR 64
+#define _RISCV_SZINT 64
+#else
+/* Our RV64 64-bit routine is equivalent to our RV32 32-bit routine.  */
+# define __muldi3 __mulsi3
+#define _RISCV_SZPTR 32
+#define _RISCV_SZINT 32
+#endif
+
+  .globl __muldi3
+__muldi3:
+  mv     a2, a0
+  li     a0, 0
+.L1:
+  slli   a3, a1, _RISCV_SZPTR-1
+  bgez   a3, .L2
+  add    a0, a0, a2
+.L2:
+  srli   a1, a1, 1
+  slli   a2, a2, 1
+  bnez   a1, .L1
+  ret

sw/src/printf.c

@@ -0,0 +1,914 @@
+///////////////////////////////////////////////////////////////////////////////
+// \author (c) Marco Paland (info@paland.com)
+//             2014-2019, PALANDesign Hannover, Germany
+//
+// \license The MIT License (MIT)
+//
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in
+// all copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+// THE SOFTWARE.
+//
+// \brief Tiny printf, sprintf and (v)snprintf implementation, optimized for speed on
+//        embedded systems with a very limited resources. These routines are thread
+//        safe and reentrant!
+//        Use this instead of the bloated standard/newlib printf cause these use
+//        malloc for printf (and may not be thread safe).
+//
+///////////////////////////////////////////////////////////////////////////////
+
+#include <stdbool.h>
+#include <stdint.h>
+
+#include "printf.h"
+
+
+// define this globally (e.g. gcc -DPRINTF_INCLUDE_CONFIG_H ...) to include the
+// printf_config.h header file
+// default: undefined
+// #ifdef PRINTF_INCLUDE_CONFIG_H
+// #include "printf_config.h"
+// #endif
+
+
+// 'ntoa' conversion buffer size, this must be big enough to hold one converted
+// numeric number including padded zeros (dynamically created on stack)
+// default: 32 byte
+#ifndef PRINTF_NTOA_BUFFER_SIZE
+#define PRINTF_NTOA_BUFFER_SIZE    32U
+#endif
+
+// 'ftoa' conversion buffer size, this must be big enough to hold one converted
+// float number including padded zeros (dynamically created on stack)
+// default: 32 byte
+#ifndef PRINTF_FTOA_BUFFER_SIZE
+#define PRINTF_FTOA_BUFFER_SIZE    32U
+#endif
+
+// support for the floating point type (%f)
+// default: activated
+// #ifndef PRINTF_DISABLE_SUPPORT_FLOAT
+// #define PRINTF_SUPPORT_FLOAT
+// #endif
+
+// support for exponential floating point notation (%e/%g)
+// default: activated
+// #ifndef PRINTF_DISABLE_SUPPORT_EXPONENTIAL
+// #define PRINTF_SUPPORT_EXPONENTIAL
+// #endif
+
+// define the default floating point precision
+// default: 6 digits
+#ifndef PRINTF_DEFAULT_FLOAT_PRECISION
+#define PRINTF_DEFAULT_FLOAT_PRECISION  6U
+#endif
+
+// define the largest float suitable to print with %f
+// default: 1e9
+#ifndef PRINTF_MAX_FLOAT
+#define PRINTF_MAX_FLOAT  1e9
+#endif
+
+// support for the long long types (%llu or %p)
+// default: activated
+// #ifndef PRINTF_DISABLE_SUPPORT_LONG_LONG
+// #define PRINTF_SUPPORT_LONG_LONG
+// #endif
+
+// support for the ptrdiff_t type (%t)
+// ptrdiff_t is normally defined in <stddef.h> as long or long long type
+// default: activated
+#ifndef PRINTF_DISABLE_SUPPORT_PTRDIFF_T
+// #define PRINTF_SUPPORT_PTRDIFF_T
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+
+// internal flag definitions
+#define FLAGS_ZEROPAD   (1U <<  0U)
+#define FLAGS_LEFT      (1U <<  1U)
+#define FLAGS_PLUS      (1U <<  2U)
+#define FLAGS_SPACE     (1U <<  3U)
+#define FLAGS_HASH      (1U <<  4U)
+#define FLAGS_UPPERCASE (1U <<  5U)
+#define FLAGS_CHAR      (1U <<  6U)
+#define FLAGS_SHORT     (1U <<  7U)
+#define FLAGS_LONG      (1U <<  8U)
+#define FLAGS_LONG_LONG (1U <<  9U)
+#define FLAGS_PRECISION (1U << 10U)
+#define FLAGS_ADAPT_EXP (1U << 11U)
+
+
+// import float.h for DBL_MAX
+#if defined(PRINTF_SUPPORT_FLOAT)
+#include <float.h>
+#endif
+
+
+// output function type
+typedef void (*out_fct_type)(char character, void* buffer, size_t idx, size_t maxlen);
+
+
+// wrapper (used as buffer) for output function type
+typedef struct {
+  void  (*fct)(char character, void* arg);
+  void* arg;
+} out_fct_wrap_type;
+
+
+// internal buffer output
+static inline void _out_buffer(char character, void* buffer, size_t idx, size_t maxlen)
+{
+  if (idx < maxlen) {
+    ((char*)buffer)[idx] = character;
+  }
+}
+
+
+// internal null output
+static inline void _out_null(char character, void* buffer, size_t idx, size_t maxlen)
+{
+  (void)character; (void)buffer; (void)idx; (void)maxlen;
+}
+
+
+// internal _putchar wrapper
+static inline void _out_char(char character, void* buffer, size_t idx, size_t maxlen)
+{
+  (void)buffer; (void)idx; (void)maxlen;
+  if (character) {
+    _putchar(character);
+  }
+}
+
+
+// internal output function wrapper
+static inline void _out_fct(char character, void* buffer, size_t idx, size_t maxlen)
+{
+  (void)idx; (void)maxlen;
+  if (character) {
+    // buffer is the output fct pointer
+    ((out_fct_wrap_type*)buffer)->fct(character, ((out_fct_wrap_type*)buffer)->arg);
+  }
+}
+
+
+// internal secure strlen
+// \return The length of the string (excluding the terminating 0) limited by 'maxsize'
+static inline unsigned int _strnlen_s(const char* str, size_t maxsize)
+{
+  const char* s;
+  for (s = str; *s && maxsize--; ++s);
+  return (unsigned int)(s - str);
+}
+
+
+// internal test if char is a digit (0-9)
+// \return true if char is a digit
+static inline bool _is_digit(char ch)
+{
+  return (ch >= '0') && (ch <= '9');
+}
+
+
+// internal ASCII string to unsigned int conversion
+static unsigned int _atoi(const char** str)
+{
+  unsigned int i = 0U;
+  while (_is_digit(**str)) {
+    i = i * 10U + (unsigned int)(*((*str)++) - '0');
+  }
+  return i;
+}
+
+
+// output the specified string in reverse, taking care of any zero-padding
+static size_t _out_rev(out_fct_type out, char* buffer, size_t idx, size_t maxlen, const char* buf, size_t len, unsigned int width, unsigned int flags)
+{
+  const size_t start_idx = idx;
+
+  // pad spaces up to given width
+  if (!(flags & FLAGS_LEFT) && !(flags & FLAGS_ZEROPAD)) {
+    for (size_t i = len; i < width; i++) {
+      out(' ', buffer, idx++, maxlen);
+    }
+  }
+
+  // reverse string
+  while (len) {
+    out(buf[--len], buffer, idx++, maxlen);
+  }
+
+  // append pad spaces up to given width
+  if (flags & FLAGS_LEFT) {
+    while (idx - start_idx < width) {
+      out(' ', buffer, idx++, maxlen);
+    }
+  }
+
+  return idx;
+}
+
+
+// internal itoa format
+static size_t _ntoa_format(out_fct_type out, char* buffer, size_t idx, size_t maxlen, char* buf, size_t len, bool negative, unsigned int base, unsigned int prec, unsigned int width, unsigned int flags)
+{
+  // pad leading zeros
+  if (!(flags & FLAGS_LEFT)) {
+    if (width && (flags & FLAGS_ZEROPAD) && (negative || (flags & (FLAGS_PLUS | FLAGS_SPACE)))) {
+      width--;
+    }
+    while ((len < prec) && (len < PRINTF_NTOA_BUFFER_SIZE)) {
+      buf[len++] = '0';
+    }
+    while ((flags & FLAGS_ZEROPAD) && (len < width) && (len < PRINTF_NTOA_BUFFER_SIZE)) {
+      buf[len++] = '0';
+    }
+  }
+
+  // handle hash
+  if (flags & FLAGS_HASH) {
+    if (!(flags & FLAGS_PRECISION) && len && ((len == prec) || (len == width))) {
+      len--;
+      if (len && (base == 16U)) {
+        len--;
+      }
+    }
+    if ((base == 16U) && !(flags & FLAGS_UPPERCASE) && (len < PRINTF_NTOA_BUFFER_SIZE)) {
+      buf[len++] = 'x';
+    }
+    else if ((base == 16U) && (flags & FLAGS_UPPERCASE) && (len < PRINTF_NTOA_BUFFER_SIZE)) {
+      buf[len++] = 'X';
+    }
+    else if ((base == 2U) && (len < PRINTF_NTOA_BUFFER_SIZE)) {
+      buf[len++] = 'b';
+    }
+    if (len < PRINTF_NTOA_BUFFER_SIZE) {
+      buf[len++] = '0';
+    }
+  }
+
+  if (len < PRINTF_NTOA_BUFFER_SIZE) {
+    if (negative) {
+      buf[len++] = '-';
+    }
+    else if (flags & FLAGS_PLUS) {
+      buf[len++] = '+';  // ignore the space if the '+' exists
+    }
+    else if (flags & FLAGS_SPACE) {
+      buf[len++] = ' ';
+    }
+  }
+
+  return _out_rev(out, buffer, idx, maxlen, buf, len, width, flags);
+}
+
+
+// internal itoa for 'long' type
+static size_t _ntoa_long(out_fct_type out, char* buffer, size_t idx, size_t maxlen, unsigned long value, bool negative, unsigned long base, unsigned int prec, unsigned int width, unsigned int flags)
+{
+  char buf[PRINTF_NTOA_BUFFER_SIZE];
+  size_t len = 0U;
+
+  // no hash for 0 values
+  if (!value) {
+    flags &= ~FLAGS_HASH;
+  }
+
+  // write if precision != 0 and value is != 0
+  if (!(flags & FLAGS_PRECISION) || value) {
+    do {
+      const char digit = (char)(value % base);
+      buf[len++] = digit < 10 ? '0' + digit : (flags & FLAGS_UPPERCASE ? 'A' : 'a') + digit - 10;
+      value /= base;
+    } while (value && (len < PRINTF_NTOA_BUFFER_SIZE));
+  }
+
+  return _ntoa_format(out, buffer, idx, maxlen, buf, len, negative, (unsigned int)base, prec, width, flags);
+}
+
+
+// internal itoa for 'long long' type
+#if defined(PRINTF_SUPPORT_LONG_LONG)
+static size_t _ntoa_long_long(out_fct_type out, char* buffer, size_t idx, size_t maxlen, unsigned long long value, bool negative, unsigned long long base, unsigned int prec, unsigned int width, unsigned int flags)
+{
+  char buf[PRINTF_NTOA_BUFFER_SIZE];
+  size_t len = 0U;
+
+  // no hash for 0 values
+  if (!value) {
+    flags &= ~FLAGS_HASH;
+  }
+
+  // write if precision != 0 and value is != 0
+  if (!(flags & FLAGS_PRECISION) || value) {
+    do {
+      const char digit = (char)(value % base);
+      buf[len++] = digit < 10 ? '0' + digit : (flags & FLAGS_UPPERCASE ? 'A' : 'a') + digit - 10;
+      value /= base;
+    } while (value && (len < PRINTF_NTOA_BUFFER_SIZE));
+  }
+
+  return _ntoa_format(out, buffer, idx, maxlen, buf, len, negative, (unsigned int)base, prec, width, flags);
+}
+#endif  // PRINTF_SUPPORT_LONG_LONG
+
+
+#if defined(PRINTF_SUPPORT_FLOAT)
+
+#if defined(PRINTF_SUPPORT_EXPONENTIAL)
+// forward declaration so that _ftoa can switch to exp notation for values > PRINTF_MAX_FLOAT
+static size_t _etoa(out_fct_type out, char* buffer, size_t idx, size_t maxlen, double value, unsigned int prec, unsigned int width, unsigned int flags);
+#endif
+
+
+// internal ftoa for fixed decimal floating point
+static size_t _ftoa(out_fct_type out, char* buffer, size_t idx, size_t maxlen, double value, unsigned int prec, unsigned int width, unsigned int flags)
+{
+  char buf[PRINTF_FTOA_BUFFER_SIZE];
+  size_t len  = 0U;
+  double diff = 0.0;
+
+  // powers of 10
+  static const double pow10[] = { 1, 10, 100, 1000, 10000, 100000, 1000000, 10000000, 100000000, 1000000000 };
+
+  // test for special values
+  if (value != value)
+    return _out_rev(out, buffer, idx, maxlen, "nan", 3, width, flags);
+  if (value < -DBL_MAX)
+    return _out_rev(out, buffer, idx, maxlen, "fni-", 4, width, flags);
+  if (value > DBL_MAX)
+    return _out_rev(out, buffer, idx, maxlen, (flags & FLAGS_PLUS) ? "fni+" : "fni", (flags & FLAGS_PLUS) ? 4U : 3U, width, flags);
+
+  // test for very large values
+  // standard printf behavior is to print EVERY whole number digit -- which could be 100s of characters overflowing your buffers == bad
+  if ((value > PRINTF_MAX_FLOAT) || (value < -PRINTF_MAX_FLOAT)) {
+#if defined(PRINTF_SUPPORT_EXPONENTIAL)
+    return _etoa(out, buffer, idx, maxlen, value, prec, width, flags);
+#else
+    return 0U;
+#endif
+  }
+
+  // test for negative
+  bool negative = false;
+  if (value < 0) {
+    negative = true;
+    value = 0 - value;
+  }
+
+  // set default precision, if not set explicitly
+  if (!(flags & FLAGS_PRECISION)) {
+    prec = PRINTF_DEFAULT_FLOAT_PRECISION;
+  }
+  // limit precision to 9, cause a prec >= 10 can lead to overflow errors
+  while ((len < PRINTF_FTOA_BUFFER_SIZE) && (prec > 9U)) {
+    buf[len++] = '0';
+    prec--;
+  }
+
+  int whole = (int)value;
+  double tmp = (value - whole) * pow10[prec];
+  unsigned long frac = (unsigned long)tmp;
+  diff = tmp - frac;
+
+  if (diff > 0.5) {
+    ++frac;
+    // handle rollover, e.g. case 0.99 with prec 1 is 1.0
+    if (frac >= pow10[prec]) {
+      frac = 0;
+      ++whole;
+    }
+  }
+  else if (diff < 0.5) {
+  }
+  else if ((frac == 0U) || (frac & 1U)) {
+    // if halfway, round up if odd OR if last digit is 0
+    ++frac;
+  }
+
+  if (prec == 0U) {
+    diff = value - (double)whole;
+    if ((!(diff < 0.5) || (diff > 0.5)) && (whole & 1)) {
+      // exactly 0.5 and ODD, then round up
+      // 1.5 -> 2, but 2.5 -> 2
+      ++whole;
+    }
+  }
+  else {
+    unsigned int count = prec;
+    // now do fractional part, as an unsigned number
+    while (len < PRINTF_FTOA_BUFFER_SIZE) {
+      --count;
+      buf[len++] = (char)(48U + (frac % 10U));
+      if (!(frac /= 10U)) {
+        break;
+      }
+    }
+    // add extra 0s
+    while ((len < PRINTF_FTOA_BUFFER_SIZE) && (count-- > 0U)) {
+      buf[len++] = '0';
+    }
+    if (len < PRINTF_FTOA_BUFFER_SIZE) {
+      // add decimal
+      buf[len++] = '.';
+    }
+  }
+
+  // do whole part, number is reversed
+  while (len < PRINTF_FTOA_BUFFER_SIZE) {
+    buf[len++] = (char)(48 + (whole % 10));
+    if (!(whole /= 10)) {
+      break;
+    }
+  }
+
+  // pad leading zeros
+  if (!(flags & FLAGS_LEFT) && (flags & FLAGS_ZEROPAD)) {
+    if (width && (negative || (flags & (FLAGS_PLUS | FLAGS_SPACE)))) {
+      width--;
+    }
+    while ((len < width) && (len < PRINTF_FTOA_BUFFER_SIZE)) {
+      buf[len++] = '0';
+    }
+  }
+
+  if (len < PRINTF_FTOA_BUFFER_SIZE) {
+    if (negative) {
+      buf[len++] = '-';
+    }
+    else if (flags & FLAGS_PLUS) {
+      buf[len++] = '+';  // ignore the space if the '+' exists
+    }
+    else if (flags & FLAGS_SPACE) {
+      buf[len++] = ' ';
+    }
+  }
+
+  return _out_rev(out, buffer, idx, maxlen, buf, len, width, flags);
+}
+
+
+#if defined(PRINTF_SUPPORT_EXPONENTIAL)
+// internal ftoa variant for exponential floating-point type, contributed by Martijn Jasperse <m.jasperse@gmail.com>
+static size_t _etoa(out_fct_type out, char* buffer, size_t idx, size_t maxlen, double value, unsigned int prec, unsigned int width, unsigned int flags)
+{
+  // check for NaN and special values
+  if ((value != value) || (value > DBL_MAX) || (value < -DBL_MAX)) {
+    return _ftoa(out, buffer, idx, maxlen, value, prec, width, flags);
+  }
+
+  // determine the sign
+  const bool negative = value < 0;
+  if (negative) {
+    value = -value;
+  }
+
+  // default precision
+  if (!(flags & FLAGS_PRECISION)) {
+    prec = PRINTF_DEFAULT_FLOAT_PRECISION;
+  }
+
+  // determine the decimal exponent
+  // based on the algorithm by David Gay (https://www.ampl.com/netlib/fp/dtoa.c)
+  union {
+    uint64_t U;
+    double   F;
+  } conv;
+
+  conv.F = value;
+  int exp2 = (int)((conv.U >> 52U) & 0x07FFU) - 1023;           // effectively log2
+  conv.U = (conv.U & ((1ULL << 52U) - 1U)) | (1023ULL << 52U);  // drop the exponent so conv.F is now in [1,2)
+  // now approximate log10 from the log2 integer part and an expansion of ln around 1.5
+  int expval = (int)(0.1760912590558 + exp2 * 0.301029995663981 + (conv.F - 1.5) * 0.289529654602168);
+  // now we want to compute 10^expval but we want to be sure it won't overflow
+  exp2 = (int)(expval * 3.321928094887362 + 0.5);
+  const double z  = expval * 2.302585092994046 - exp2 * 0.6931471805599453;
+  const double z2 = z * z;
+  conv.U = (uint64_t)(exp2 + 1023) << 52U;
+  // compute exp(z) using continued fractions, see https://en.wikipedia.org/wiki/Exponential_function#Continued_fractions_for_ex
+  conv.F *= 1 + 2 * z / (2 - z + (z2 / (6 + (z2 / (10 + z2 / 14)))));
+  // correct for rounding errors
+  if (value < conv.F) {
+    expval--;
+    conv.F /= 10;
+  }
+
+  // the exponent format is "%+03d" and largest value is "307", so set aside 4-5 characters
+  unsigned int minwidth = ((expval < 100) && (expval > -100)) ? 4U : 5U;
+
+  // in "%g" mode, "prec" is the number of *significant figures* not decimals
+  if (flags & FLAGS_ADAPT_EXP) {
+    // do we want to fall-back to "%f" mode?
+    if ((value >= 1e-4) && (value < 1e6)) {
+      if ((int)prec > expval) {
+        prec = (unsigned)((int)prec - expval - 1);
+      }
+      else {
+        prec = 0;
+      }
+      flags |= FLAGS_PRECISION;   // make sure _ftoa respects precision
+      // no characters in exponent
+      minwidth = 0U;
+      expval   = 0;
+    }
+    else {
+      // we use one sigfig for the whole part
+      if ((prec > 0) && (flags & FLAGS_PRECISION)) {
+        --prec;
+      }
+    }
+  }
+
+  // will everything fit?
+  unsigned int fwidth = width;
+  if (width > minwidth) {
+    // we didn't fall-back so subtract the characters required for the exponent
+    fwidth -= minwidth;
+  } else {
+    // not enough characters, so go back to default sizing
+    fwidth = 0U;
+  }
+  if ((flags & FLAGS_LEFT) && minwidth) {
+    // if we're padding on the right, DON'T pad the floating part
+    fwidth = 0U;
+  }
+
+  // rescale the float value
+  if (expval) {
+    value /= conv.F;
+  }
+
+  // output the floating part
+  const size_t start_idx = idx;
+  idx = _ftoa(out, buffer, idx, maxlen, negative ? -value : value, prec, fwidth, flags & ~FLAGS_ADAPT_EXP);
+
+  // output the exponent part
+  if (minwidth) {
+    // output the exponential symbol
+    out((flags & FLAGS_UPPERCASE) ? 'E' : 'e', buffer, idx++, maxlen);
+    // output the exponent value
+    idx = _ntoa_long(out, buffer, idx, maxlen, (expval < 0) ? -expval : expval, expval < 0, 10, 0, minwidth-1, FLAGS_ZEROPAD | FLAGS_PLUS);
+    // might need to right-pad spaces
+    if (flags & FLAGS_LEFT) {
+      while (idx - start_idx < width) out(' ', buffer, idx++, maxlen);
+    }
+  }
+  return idx;
+}
+#endif  // PRINTF_SUPPORT_EXPONENTIAL
+#endif  // PRINTF_SUPPORT_FLOAT
+
+
+// internal vsnprintf
+static int _vsnprintf(out_fct_type out, char* buffer, const size_t maxlen, const char* format, va_list va)
+{
+  unsigned int flags, width, precision, n;
+  size_t idx = 0U;
+
+  if (!buffer) {
+    // use null output function
+    out = _out_null;
+  }
+
+  while (*format)
+  {
+    // format specifier?  %[flags][width][.precision][length]
+    if (*format != '%') {
+      // no
+      out(*format, buffer, idx++, maxlen);
+      format++;
+      continue;
+    }
+    else {
+      // yes, evaluate it
+      format++;
+    }
+
+    // evaluate flags
+    flags = 0U;
+    do {
+      switch (*format) {
+        case '0': flags |= FLAGS_ZEROPAD; format++; n = 1U; break;
+        case '-': flags |= FLAGS_LEFT;    format++; n = 1U; break;
+        case '+': flags |= FLAGS_PLUS;    format++; n = 1U; break;
+        case ' ': flags |= FLAGS_SPACE;   format++; n = 1U; break;
+        case '#': flags |= FLAGS_HASH;    format++; n = 1U; break;
+        default :                                   n = 0U; break;
+      }
+    } while (n);
+
+    // evaluate width field
+    width = 0U;
+    if (_is_digit(*format)) {
+      width = _atoi(&format);
+    }
+    else if (*format == '*') {
+      const int w = va_arg(va, int);
+      if (w < 0) {
+        flags |= FLAGS_LEFT;    // reverse padding
+        width = (unsigned int)-w;
+      }
+      else {
+        width = (unsigned int)w;
+      }
+      format++;
+    }
+
+    // evaluate precision field
+    precision = 0U;
+    if (*format == '.') {
+      flags |= FLAGS_PRECISION;
+      format++;
+      if (_is_digit(*format)) {
+        precision = _atoi(&format);
+      }
+      else if (*format == '*') {
+        const int prec = (int)va_arg(va, int);
+        precision = prec > 0 ? (unsigned int)prec : 0U;
+        format++;
+      }
+    }
+
+    // evaluate length field
+    switch (*format) {
+      case 'l' :
+        flags |= FLAGS_LONG;
+        format++;
+        if (*format == 'l') {
+          flags |= FLAGS_LONG_LONG;
+          format++;
+        }
+        break;
+      case 'h' :
+        flags |= FLAGS_SHORT;
+        format++;
+        if (*format == 'h') {
+          flags |= FLAGS_CHAR;
+          format++;
+        }
+        break;
+#if defined(PRINTF_SUPPORT_PTRDIFF_T)
+      case 't' :
+        flags |= (sizeof(ptrdiff_t) == sizeof(long) ? FLAGS_LONG : FLAGS_LONG_LONG);
+        format++;
+        break;
+#endif
+      case 'j' :
+        flags |= (sizeof(intmax_t) == sizeof(long) ? FLAGS_LONG : FLAGS_LONG_LONG);
+        format++;
+        break;
+      case 'z' :
+        flags |= (sizeof(size_t) == sizeof(long) ? FLAGS_LONG : FLAGS_LONG_LONG);
+        format++;
+        break;
+      default :
+        break;
+    }
+
+    // evaluate specifier
+    switch (*format) {
+      case 'd' :
+      case 'i' :
+      case 'u' :
+      case 'x' :
+      case 'X' :
+      case 'o' :
+      case 'b' : {
+        // set the base
+        unsigned int base;
+        if (*format == 'x' || *format == 'X') {
+          base = 16U;
+        }
+        else if (*format == 'o') {
+          base =  8U;
+        }
+        else if (*format == 'b') {
+          base =  2U;
+        }
+        else {
+          base = 10U;
+          flags &= ~FLAGS_HASH;   // no hash for dec format
+        }
+        // uppercase
+        if (*format == 'X') {
+          flags |= FLAGS_UPPERCASE;
+        }
+
+        // no plus or space flag for u, x, X, o, b
+        if ((*format != 'i') && (*format != 'd')) {
+          flags &= ~(FLAGS_PLUS | FLAGS_SPACE);
+        }
+
+        // ignore '0' flag when precision is given
+        if (flags & FLAGS_PRECISION) {
+          flags &= ~FLAGS_ZEROPAD;
+        }
+
+        // convert the integer
+        if ((*format == 'i') || (*format == 'd')) {
+          // signed
+          if (flags & FLAGS_LONG_LONG) {
+#if defined(PRINTF_SUPPORT_LONG_LONG)
+            const long long value = va_arg(va, long long);
+            idx = _ntoa_long_long(out, buffer, idx, maxlen, (unsigned long long)(value > 0 ? value : 0 - value), value < 0, base, precision, width, flags);
+#endif
+          }
+          else if (flags & FLAGS_LONG) {
+            const long value = va_arg(va, long);
+            idx = _ntoa_long(out, buffer, idx, maxlen, (unsigned long)(value > 0 ? value : 0 - value), value < 0, base, precision, width, flags);
+          }
+          else {
+            const int value = (flags & FLAGS_CHAR) ? (char)va_arg(va, int) : (flags & FLAGS_SHORT) ? (short int)va_arg(va, int) : va_arg(va, int);
+            idx = _ntoa_long(out, buffer, idx, maxlen, (unsigned int)(value > 0 ? value : 0 - value), value < 0, base, precision, width, flags);
+          }
+        }
+        else {
+          // unsigned
+          if (flags & FLAGS_LONG_LONG) {
+#if defined(PRINTF_SUPPORT_LONG_LONG)
+            idx = _ntoa_long_long(out, buffer, idx, maxlen, va_arg(va, unsigned long long), false, base, precision, width, flags);
+#endif
+          }
+          else if (flags & FLAGS_LONG) {
+            idx = _ntoa_long(out, buffer, idx, maxlen, va_arg(va, unsigned long), false, base, precision, width, flags);
+          }
+          else {
+            const unsigned int value = (flags & FLAGS_CHAR) ? (unsigned char)va_arg(va, unsigned int) : (flags & FLAGS_SHORT) ? (unsigned short int)va_arg(va, unsigned int) : va_arg(va, unsigned int);
+            idx = _ntoa_long(out, buffer, idx, maxlen, value, false, base, precision, width, flags);
+          }
+        }
+        format++;
+        break;
+      }
+#if defined(PRINTF_SUPPORT_FLOAT)
+      case 'f' :
+      case 'F' :
+        if (*format == 'F') flags |= FLAGS_UPPERCASE;
+        idx = _ftoa(out, buffer, idx, maxlen, va_arg(va, double), precision, width, flags);
+        format++;
+        break;
+#if defined(PRINTF_SUPPORT_EXPONENTIAL)
+      case 'e':
+      case 'E':
+      case 'g':
+      case 'G':
+        if ((*format == 'g')||(*format == 'G')) flags |= FLAGS_ADAPT_EXP;
+        if ((*format == 'E')||(*format == 'G')) flags |= FLAGS_UPPERCASE;
+        idx = _etoa(out, buffer, idx, maxlen, va_arg(va, double), precision, width, flags);
+        format++;
+        break;
+#endif  // PRINTF_SUPPORT_EXPONENTIAL
+#endif  // PRINTF_SUPPORT_FLOAT
+      case 'c' : {
+        unsigned int l = 1U;
+        // pre padding
+        if (!(flags & FLAGS_LEFT)) {
+          while (l++ < width) {
+            out(' ', buffer, idx++, maxlen);
+          }
+        }
+        // char output
+        out((char)va_arg(va, int), buffer, idx++, maxlen);
+        // post padding
+        if (flags & FLAGS_LEFT) {
+          while (l++ < width) {
+            out(' ', buffer, idx++, maxlen);
+          }
+        }
+        format++;
+        break;
+      }
+
+      case 's' : {
+        const char* p = va_arg(va, char*);
+        unsigned int l = _strnlen_s(p, precision ? precision : (size_t)-1);
+        // pre padding
+        if (flags & FLAGS_PRECISION) {
+          l = (l < precision ? l : precision);
+        }
+        if (!(flags & FLAGS_LEFT)) {
+          while (l++ < width) {
+            out(' ', buffer, idx++, maxlen);
+          }
+        }
+        // string output
+        while ((*p != 0) && (!(flags & FLAGS_PRECISION) || precision--)) {
+          out(*(p++), buffer, idx++, maxlen);
+        }
+        // post padding
+        if (flags & FLAGS_LEFT) {
+          while (l++ < width) {
+            out(' ', buffer, idx++, maxlen);
+          }
+        }
+        format++;
+        break;
+      }
+
+      case 'p' : {
+        width = sizeof(void*) * 2U;
+        flags |= FLAGS_ZEROPAD | FLAGS_UPPERCASE;
+#if defined(PRINTF_SUPPORT_LONG_LONG)
+        const bool is_ll = sizeof(uintptr_t) == sizeof(long long);
+        if (is_ll) {
+          idx = _ntoa_long_long(out, buffer, idx, maxlen, (uintptr_t)va_arg(va, void*), false, 16U, precision, width, flags);
+        }
+        else {
+#endif
+          idx = _ntoa_long(out, buffer, idx, maxlen, (unsigned long)((uintptr_t)va_arg(va, void*)), false, 16U, precision, width, flags);
+#if defined(PRINTF_SUPPORT_LONG_LONG)
+        }
+#endif
+        format++;
+        break;
+      }
+
+      case '%' :
+        out('%', buffer, idx++, maxlen);
+        format++;
+        break;
+
+      default :
+        out(*format, buffer, idx++, maxlen);
+        format++;
+        break;
+    }
+  }
+
+  // termination
+  out((char)0, buffer, idx < maxlen ? idx : maxlen - 1U, maxlen);
+
+  // return written chars without terminating \0
+  return (int)idx;
+}
+
+
+///////////////////////////////////////////////////////////////////////////////
+
+int printf_(const char* format, ...)
+{
+  va_list va;
+  va_start(va, format);
+  char buffer[1];
+  const int ret = _vsnprintf(_out_char, buffer, (size_t)-1, format, va);
+  va_end(va);
+  return ret;
+}
+
+
+int sprintf_(char* buffer, const char* format, ...)
+{
+  va_list va;
+  va_start(va, format);
+  const int ret = _vsnprintf(_out_buffer, buffer, (size_t)-1, format, va);
+  va_end(va);
+  return ret;
+}
+
+
+int snprintf_(char* buffer, size_t count, const char* format, ...)
+{
+  va_list va;
+  va_start(va, format);
+  const int ret = _vsnprintf(_out_buffer, buffer, count, format, va);
+  va_end(va);
+  return ret;
+}
+
+
+int vprintf_(const char* format, va_list va)
+{
+  char buffer[1];
+  return _vsnprintf(_out_char, buffer, (size_t)-1, format, va);
+}
+
+
+int vsnprintf_(char* buffer, size_t count, const char* format, va_list va)
+{
+  return _vsnprintf(_out_buffer, buffer, count, format, va);
+}
+
+
+int fctprintf(void (*out)(char character, void* arg), void* arg, const char* format, ...)
+{
+  va_list va;
+  va_start(va, format);
+  const out_fct_wrap_type out_fct_wrap = { out, arg };
+  const int ret = _vsnprintf(_out_fct, (char*)(uintptr_t)&out_fct_wrap, (size_t)-1, format, va);
+  va_end(va);
+  return ret;
+}

sw/src/tester.c

@@ -1,5 +1,7 @@
 #include <tester.h>
+#include <printf.h>
 
 void tester_poll(void) {
+    printf("Hello, world!\n");
     return;
 }

sw/src/usb-cdc.c

@@ -1,4 +1,6 @@
+#include <usb.h>
 #include <usb-cdc.h>
+#include <fomu/csr.h>
 
 static int connected = 0;
 
@@ -11,3 +13,12 @@ void cdc_set_connected(int is_connected)
 {
     connected = is_connected;
 }
+
+void _putchar(char character)
+{
+    // Wait for buffer to be empty
+    while (usb_ep_2_in_respond_read() == EPF_ACK)
+        ;
+    usb_ep_2_in_ibuf_head_write(character);
+    usb_ep_2_in_respond_write(EPF_ACK);
+}

sw/src/usb-epfifo.c

@@ -5,14 +5,23 @@
 
 #ifdef CSR_USB_EP_0_OUT_EV_PENDING_ADDR
 
+static const int max_byte_length = 64;
+
 #define EP0OUT_BUFFERS 4
+#define EP2OUT_BUFFERS 4
 __attribute__((aligned(4)))
+#define EP0OUT_BUFFER_SIZE 256
 static uint8_t volatile usb_ep0out_buffer_len[EP0OUT_BUFFERS];
-static uint8_t volatile usb_ep0out_buffer[EP0OUT_BUFFERS][256];
+static uint8_t volatile usb_ep0out_buffer[EP0OUT_BUFFERS][EP0OUT_BUFFER_SIZE];
 static uint8_t volatile usb_ep0out_last_tok[EP0OUT_BUFFERS];
 static volatile uint8_t usb_ep0out_wr_ptr;
 static volatile uint8_t usb_ep0out_rd_ptr;
-static const int max_byte_length = 64;
+
+#define EP2OUT_BUFFER_SIZE 256
+static uint8_t volatile usb_ep2out_buffer_len[EP2OUT_BUFFERS];
+static uint8_t volatile usb_ep2out_buffer[EP2OUT_BUFFERS][EP2OUT_BUFFER_SIZE];
+static volatile uint8_t usb_ep2out_wr_ptr;
+static volatile uint8_t usb_ep2out_rd_ptr;
 
 static const uint8_t * volatile current_data;
 static volatile int current_length;
@@ -20,25 +29,6 @@ static volatile int data_offset;
 static volatile int data_to_send;
 static int next_packet_is_empty;
 
-// Note that our PIDs are only bits 2 and 3 of the token,
-// since all other bits are effectively redundant at this point.
-enum USB_PID {
-    USB_PID_OUT   = 0,
-    USB_PID_SOF   = 1,
-    USB_PID_IN    = 2,
-    USB_PID_SETUP = 3,
-};
-
-enum epfifo_response {
-    EPF_ACK = 0,
-    EPF_NAK = 1,
-    EPF_NONE = 2,
-    EPF_STALL = 3,
-};
-
-#define USB_EV_ERROR 1
-#define USB_EV_PACKET 2
-
 void usb_idle(void) {
     usb_ep_0_out_ev_enable_write(0);
     usb_ep_0_in_ev_enable_write(0);
@@ -60,14 +50,22 @@ void usb_disconnect(void) {
 }
 
 void usb_connect(void) {
-
     usb_ep_0_out_ev_pending_write(usb_ep_0_out_ev_enable_read());
     usb_ep_0_in_ev_pending_write(usb_ep_0_in_ev_pending_read());
     usb_ep_0_out_ev_enable_write(USB_EV_PACKET | USB_EV_ERROR);
     usb_ep_0_in_ev_enable_write(USB_EV_PACKET | USB_EV_ERROR);
 
+    usb_ep_1_in_ev_pending_write(usb_ep_1_in_ev_enable_read());
+    usb_ep_1_in_ev_enable_write(USB_EV_PACKET | USB_EV_ERROR);
+
+    usb_ep_2_out_ev_pending_write(usb_ep_2_out_ev_enable_read());
+    usb_ep_2_in_ev_pending_write(usb_ep_2_in_ev_pending_read());
+    usb_ep_2_out_ev_enable_write(USB_EV_PACKET | USB_EV_ERROR);
+    usb_ep_2_in_ev_enable_write(USB_EV_PACKET | USB_EV_ERROR);
+
     // Accept incoming data by default.
     usb_ep_0_out_respond_write(EPF_ACK);
+    usb_ep_2_out_respond_write(EPF_ACK);
 
     // Reject outgoing data, since we have none to give yet.
     usb_ep_0_in_respond_write(EPF_NAK);
@@ -153,24 +151,24 @@ void usb_wait_for_send_done(void) {
 }
 
 void usb_isr(void) {
-    uint8_t ep0o_pending = usb_ep_0_out_ev_pending_read();
+    uint8_t ep0out_pending = usb_ep_0_out_ev_pending_read();
-    uint8_t ep0i_pending = usb_ep_0_in_ev_pending_read();
+    uint8_t ep0in_pending = usb_ep_0_in_ev_pending_read();
+    uint8_t ep1in_pending = usb_ep_1_in_ev_pending_read();
+    uint8_t ep2in_pending = usb_ep_2_in_ev_pending_read();
+    uint8_t ep2out_pending = usb_ep_2_out_ev_pending_read();
 
     // We got an OUT or a SETUP packet.  Copy it to usb_ep0out_buffer
     // and clear the "pending" bit.
-    if (ep0o_pending) {
+    if (ep0out_pending) {
         uint8_t last_tok = usb_ep_0_out_last_tok_read();
         
         int byte_count = 0;
         usb_ep0out_last_tok[usb_ep0out_wr_ptr] = last_tok;
         volatile uint8_t * obuf = usb_ep0out_buffer[usb_ep0out_wr_ptr];
-        if (!usb_ep_0_out_obuf_empty_read()) {
+        while (!usb_ep_0_out_obuf_empty_read()) {
-            while (!usb_ep_0_out_obuf_empty_read()) {
+            obuf[byte_count++] = usb_ep_0_out_obuf_head_read();
-                obuf[byte_count++] = usb_ep_0_out_obuf_head_read();
+            usb_ep_0_out_obuf_head_write(0);
-                usb_ep_0_out_obuf_head_write(0);
-            }
         }
-
         if (byte_count >= 2)
             usb_ep0out_buffer_len[usb_ep0out_wr_ptr] = byte_count - 2 /* Strip off CRC16 */;
         usb_ep0out_wr_ptr = (usb_ep0out_wr_ptr + 1) & (EP0OUT_BUFFERS-1);
@@ -181,15 +179,41 @@ void usb_isr(void) {
             current_length = 0;
             current_data = NULL;
         }
-
+        usb_ep_0_out_ev_pending_write(ep0out_pending);
-        usb_ep_0_out_ev_pending_write(ep0o_pending);
         usb_ep_0_out_respond_write(EPF_ACK);
     }
 
     // We just got an "IN" token.  Send data if we have it.
-    if (ep0i_pending) {
+    if (ep0in_pending) {
         usb_ep_0_in_respond_write(EPF_NAK);
-        usb_ep_0_in_ev_pending_write(ep0i_pending);
+        usb_ep_0_in_ev_pending_write(ep0in_pending);
+    }
+
+    if (ep1in_pending) {
+        usb_ep_1_in_respond_write(EPF_NAK);
+        usb_ep_1_in_ev_pending_write(ep1in_pending);
+    }
+
+    if (ep2in_pending) {
+        usb_ep_2_in_respond_write(EPF_NAK);
+        usb_ep_2_in_ev_pending_write(ep2in_pending);
+    }
+
+    if (ep2out_pending) {
+        volatile uint8_t * obuf = usb_ep2out_buffer[usb_ep2out_wr_ptr];
+        int sz = 0;
+        while (!usb_ep_2_out_obuf_empty_read()) {
+            if (sz < EP2OUT_BUFFER_SIZE)
+                obuf[sz++] = usb_ep_2_out_obuf_head_read() + 1;
+            usb_ep_2_out_obuf_head_write(0);
+        }
+        if (sz > 2) {
+            usb_ep2out_buffer_len[usb_ep2out_wr_ptr] = sz - 2; /* Strip off CRC16 */
+            usb_ep2out_wr_ptr = (usb_ep2out_wr_ptr + 1) & (EP2OUT_BUFFERS-1);
+        }
+
+        usb_ep_2_out_respond_write(EPF_ACK);
+        usb_ep_2_out_ev_pending_write(ep2out_pending);
     }
 
     return;