zy1000: print out khz correctly in response to setting JTAG speed
[openocd-genbsdl] / src / jtag / zy1000 / zy1000.c
index f4fee3a..23f2fdd 100644 (file)
@@ -1,5 +1,5 @@
-/***************************************************************************
- *   Copyright (C) 2007-2008 by Øyvind Harboe                              *
+'n/***************************************************************************
+ *   Copyright (C) 2007-2010 by Øyvind Harboe                              *
  *                                                                         *
  *   This program is free software; you can redistribute it and/or modify  *
  *   it under the terms of the GNU General Public License as published by  *
  *   Free Software Foundation, Inc.,                                       *
  *   59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.             *
  ***************************************************************************/
+
+/* This file supports the zy1000 debugger: http://www.zylin.com/zy1000.html
+ *
+ * The zy1000 is a standalone debugger that has a web interface and
+ * requires no drivers on the developer host as all communication
+ * is via TCP/IP. The zy1000 gets it performance(~400-700kBytes/s
+ * DCC downloads @ 16MHz target) as it has an FPGA to hardware
+ * accelerate the JTAG commands, while offering *very* low latency
+ * between OpenOCD and the FPGA registers.
+ *
+ * The disadvantage of the zy1000 is that it has a feeble CPU compared to
+ * a PC(ca. 50-500 DMIPS depending on how one counts it), whereas a PC
+ * is on the order of 10000 DMIPS(i.e. at a factor of 20-200).
+ *
+ * The zy1000 revc hardware is using an Altera Nios CPU, whereas the
+ * revb is using ARM7 + Xilinx.
+ *
+ * See Zylin web pages or contact Zylin for more information.
+ *
+ * The reason this code is in OpenOCD rather than OpenOCD linked with the
+ * ZY1000 code is that OpenOCD is the long road towards getting
+ * libopenocd into place. libopenocd will support both low performance,
+ * low latency systems(embedded) and high performance high latency
+ * systems(PCs).
+ */
 #ifdef HAVE_CONFIG_H
 #include "config.h"
 #endif
 
-#include "embeddedice.h"
-#include "minidriver.h"
-#include "interface.h"
+#include <target/embeddedice.h>
+#include <jtag/minidriver.h>
+#include <jtag/interface.h>
+#include <time.h>
+#include <helper/time_support.h>
+
+#include <netinet/tcp.h>
+
+#if BUILD_ECOSBOARD
+#include "zy1000_version.h"
 
 #include <cyg/hal/hal_io.h>             // low level i/o
 #include <cyg/hal/hal_diag.h>
 
+#ifdef CYGPKG_HAL_NIOS2
+#include <cyg/hal/io.h>
+#include <cyg/firmwareutil/firmwareutil.h>
+#endif
 
-#define ZYLIN_VERSION "1.52"
+#define ZYLIN_VERSION GIT_ZY1000_VERSION
 #define ZYLIN_DATE __DATE__
 #define ZYLIN_TIME __TIME__
-#define ZYLIN_OPENOCD "$Revision$"
-#define ZYLIN_OPENOCD_VERSION "Zylin JTAG ZY1000 " ZYLIN_VERSION " " ZYLIN_DATE " " ZYLIN_TIME
-
-/* low level command set
- */
-void zy1000_reset(int trst, int srst);
+#define ZYLIN_OPENOCD GIT_OPENOCD_VERSION
+#define ZYLIN_OPENOCD_VERSION "ZY1000 " ZYLIN_VERSION " " ZYLIN_DATE
 
+#endif
 
-int zy1000_speed(int speed);
-int zy1000_register_commands(struct command_context_s *cmd_ctx);
-int zy1000_init(void);
-int zy1000_quit(void);
 
-/* interface commands */
-int zy1000_handle_zy1000_port_command(struct command_context_s *cmd_ctx, char *cmd, char **args, int argc);
+/* The software needs to check if it's in RCLK mode or not */
+static bool zy1000_rclk = false;
 
 static int zy1000_khz(int khz, int *jtag_speed)
 {
-       if (khz==0)
+       if (khz == 0)
        {
-               *jtag_speed=0;
+               *jtag_speed = 0;
        }
        else
        {
-               *jtag_speed=64000/khz;
+               int speed;
+               /* Round speed up to nearest divisor.
+                *
+                * E.g. 16000kHz
+                * (64000 + 15999) / 16000 = 4
+                * (4 + 1) / 2 = 2
+                * 2 * 2 = 4
+                *
+                * 64000 / 4 = 16000
+                *
+                * E.g. 15999
+                * (64000 + 15998) / 15999 = 5
+                * (5 + 1) / 2 = 3
+                * 3 * 2 = 6
+                *
+                * 64000 / 6 = 10666
+                *
+                */
+               speed = (64000 + (khz -1)) / khz;
+               speed = (speed + 1 ) / 2;
+               speed *= 2;
+               if (speed > 8190)
+               {
+                       /* maximum dividend */
+                       speed = 8190;
+               }
+               *jtag_speed = speed;
        }
        return ERROR_OK;
 }
 
 static int zy1000_speed_div(int speed, int *khz)
 {
-       if (speed==0)
+       if (speed == 0)
        {
                *khz = 0;
        }
        else
        {
-               *khz=64000/speed;
+               *khz = 64000/speed;
        }
 
        return ERROR_OK;
@@ -76,10 +131,10 @@ static int zy1000_speed_div(int speed, int *khz)
 
 static bool readPowerDropout(void)
 {
-       cyg_uint32 state;
+       uint32_t state;
        // sample and clear power dropout
-       HAL_WRITE_UINT32(ZY1000_JTAG_BASE+0x10, 0x80);
-       HAL_READ_UINT32(ZY1000_JTAG_BASE+0x10, state);
+       ZY1000_POKE(ZY1000_JTAG_BASE + 0x10, 0x80);
+       ZY1000_PEEK(ZY1000_JTAG_BASE + 0x10, state);
        bool powerDropout;
        powerDropout = (state & 0x80) != 0;
        return powerDropout;
@@ -88,10 +143,10 @@ static bool readPowerDropout(void)
 
 static bool readSRST(void)
 {
-       cyg_uint32 state;
+       uint32_t state;
        // sample and clear SRST sensing
-       HAL_WRITE_UINT32(ZY1000_JTAG_BASE+0x10, 0x00000040);
-       HAL_READ_UINT32(ZY1000_JTAG_BASE+0x10, state);
+       ZY1000_POKE(ZY1000_JTAG_BASE + 0x10, 0x00000040);
+       ZY1000_PEEK(ZY1000_JTAG_BASE + 0x10, state);
        bool srstAsserted;
        srstAsserted = (state & 0x40) != 0;
        return srstAsserted;
@@ -99,113 +154,127 @@ static bool readSRST(void)
 
 static int zy1000_srst_asserted(int *srst_asserted)
 {
-       *srst_asserted=readSRST();
+       *srst_asserted = readSRST();
        return ERROR_OK;
 }
 
 static int zy1000_power_dropout(int *dropout)
 {
-       *dropout=readPowerDropout();
+       *dropout = readPowerDropout();
        return ERROR_OK;
 }
 
-
-jtag_interface_t zy1000_interface =
-{
-       .name = "ZY1000",
-       .execute_queue = NULL,
-       .speed = zy1000_speed,
-       .register_commands = zy1000_register_commands,
-       .init = zy1000_init,
-       .quit = zy1000_quit,
-       .khz = zy1000_khz,
-       .speed_div = zy1000_speed_div,
-       .power_dropout = zy1000_power_dropout,
-       .srst_asserted = zy1000_srst_asserted,
-};
-
 void zy1000_reset(int trst, int srst)
 {
        LOG_DEBUG("zy1000 trst=%d, srst=%d", trst, srst);
-       if(!srst)
+
+       /* flush the JTAG FIFO. Not flushing the queue before messing with
+        * reset has such interesting bugs as causing hard to reproduce
+        * RCLK bugs as RCLK will stop responding when TRST is asserted
+        */
+       waitIdle();
+
+       if (!srst)
        {
-               ZY1000_POKE(ZY1000_JTAG_BASE+0x14, 0x00000001);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x14, 0x00000001);
        }
        else
        {
-               /* Danger!!! if clk!=0 when in
+               /* Danger!!! if clk != 0 when in
                 * idle in TAP_IDLE, reset halt on str912 will fail.
                 */
-               ZY1000_POKE(ZY1000_JTAG_BASE+0x10, 0x00000001);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x10, 0x00000001);
        }
 
-       if(!trst)
+       if (!trst)
        {
-               ZY1000_POKE(ZY1000_JTAG_BASE+0x14, 0x00000002);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x14, 0x00000002);
        }
        else
        {
                /* assert reset */
-               ZY1000_POKE(ZY1000_JTAG_BASE+0x10, 0x00000002);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x10, 0x00000002);
        }
 
-       if (trst||(srst&&(jtag_get_reset_config() & RESET_SRST_PULLS_TRST)))
+       if (trst||(srst && (jtag_get_reset_config() & RESET_SRST_PULLS_TRST)))
        {
-               waitIdle();
                /* we're now in the RESET state until trst is deasserted */
-               ZY1000_POKE(ZY1000_JTAG_BASE+0x20, TAP_RESET);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x20, TAP_RESET);
        } else
        {
                /* We'll get RCLK failure when we assert TRST, so clear any false positives here */
-               ZY1000_POKE(ZY1000_JTAG_BASE+0x14, 0x400);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x14, 0x400);
        }
 
        /* wait for srst to float back up */
-       if (!srst)
+       if ((!srst && ((jtag_get_reset_config() & RESET_TRST_PULLS_SRST) == 0))||
+               (!srst && !trst && (jtag_get_reset_config() & RESET_TRST_PULLS_SRST)))
        {
-               int i;
-               for (i=0; i<1000; i++)
-               {
+               bool first = true;
+               long long start = 0;
+               long total = 0;
+               for (;;)
+               {       
                        // We don't want to sense our own reset, so we clear here.
                        // There is of course a timing hole where we could loose
                        // a "real" reset.
                        if (!readSRST())
+                       {
+                               if (total > 1)
+                               {
+                                 LOG_USER("SRST took %dms to deassert", (int)total);
+                               }
                                break;
+                       }
 
-                       /* wait 1ms */
-                       alive_sleep(1);
-               }
+                       if (first)
+                       {
+                           first = false;
+                           start = timeval_ms();
+                       }
 
-               if (i==1000)
-               {
-                       LOG_USER("SRST didn't deassert after %dms", i);
-               } else if (i>1)
-               {
-                       LOG_USER("SRST took %dms to deassert", i);
+                       total = timeval_ms() - start;
+
+                       keep_alive();
+
+                       if (total > 5000)
+                       {
+                               LOG_ERROR("SRST took too long to deassert: %dms", (int)total);
+                           break;
+                       }
                }
+
        }
 }
 
 int zy1000_speed(int speed)
 {
-       if(speed == 0)
+       /* flush JTAG master FIFO before setting speed */
+       waitIdle();
+
+       zy1000_rclk = false;
+
+       if (speed == 0)
        {
                /*0 means RCLK*/
-               speed = 0;
-               ZY1000_POKE(ZY1000_JTAG_BASE+0x10, 0x100);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x10, 0x100);
+               zy1000_rclk = true;
                LOG_DEBUG("jtag_speed using RCLK");
        }
        else
        {
-               if(speed > 8190 || speed < 2)
+               if (speed > 8190 || speed < 2)
                {
                        LOG_USER("valid ZY1000 jtag_speed=[8190,2]. Divisor is 64MHz / even values between 8190-2, i.e. min 7814Hz, max 32MHz");
                        return ERROR_INVALID_ARGUMENTS;
                }
 
-               LOG_USER("jtag_speed %d => JTAG clk=%f", speed, 64.0/(float)speed);
-               ZY1000_POKE(ZY1000_JTAG_BASE+0x14, 0x100);
-               ZY1000_POKE(ZY1000_JTAG_BASE+0x1c, speed&~1);
+               int khz;
+               speed &= ~1;
+               zy1000_speed_div(speed, &khz);
+               LOG_USER("jtag_speed %d => JTAG clk=%d kHz", speed, khz);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x14, 0x100);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x1c, speed);
        }
        return ERROR_OK;
 }
@@ -218,71 +287,110 @@ static void setPower(bool power)
        savePower = power;
        if (power)
        {
-               HAL_WRITE_UINT32(ZY1000_JTAG_BASE+0x14, 0x8);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x14, 0x8);
        } else
        {
-               HAL_WRITE_UINT32(ZY1000_JTAG_BASE+0x10, 0x8);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x10, 0x8);
        }
 }
 
-int handle_power_command(struct command_context_s *cmd_ctx, char *cmd, char **args, int argc)
+COMMAND_HANDLER(handle_power_command)
 {
-       if (argc > 1)
+       switch (CMD_ARGC)
        {
-               return ERROR_INVALID_ARGUMENTS;
+       case 1: {
+               bool enable;
+               COMMAND_PARSE_ON_OFF(CMD_ARGV[0], enable);
+               setPower(enable);
+               // fall through
        }
-
-       if (argc == 1)
-       {
-               if (strcmp(args[0], "on") == 0)
-               {
-                       setPower(1);
-               }
-               else if (strcmp(args[0], "off") == 0)
-               {
-                       setPower(0);
-               } else
-               {
-                       command_print(cmd_ctx, "arg is \"on\" or \"off\"");
-                       return ERROR_INVALID_ARGUMENTS;
-               }
+       case 0:
+               LOG_INFO("Target power %s", savePower ? "on" : "off");
+               break;
+       default:
+               return ERROR_INVALID_ARGUMENTS;
        }
 
-       command_print(cmd_ctx, "Target power %s", savePower ? "on" : "off");
-
        return ERROR_OK;
 }
 
+#if !BUILD_ECOSBOARD
+static char *tcp_server = "notspecified";
+static int jim_zy1000_server(Jim_Interp *interp, int argc, Jim_Obj *const *argv)
+{
+       if (argc != 2)
+               return JIM_ERR;
+
+       tcp_server = strdup(Jim_GetString(argv[1], NULL));
+
+       return JIM_OK;
+}
+#endif
 
+#if BUILD_ECOSBOARD
 /* Give TELNET a way to find out what version this is */
 static int jim_zy1000_version(Jim_Interp *interp, int argc, Jim_Obj *const *argv)
 {
-       if ((argc < 1) || (argc > 2))
+       if ((argc < 1) || (argc > 3))
                return JIM_ERR;
-       char buff[128];
-       const char *version_str=NULL;
+       const char *version_str = NULL;
 
        if (argc == 1)
        {
-               version_str=ZYLIN_OPENOCD_VERSION;
+               version_str = ZYLIN_OPENOCD_VERSION;
        } else
        {
                const char *str = Jim_GetString(argv[1], NULL);
+               const char *str2 = NULL;
+               if (argc > 2)
+                       str2 = Jim_GetString(argv[2], NULL);
                if (strcmp("openocd", str) == 0)
                {
-                       int revision;
-                       revision = atol(ZYLIN_OPENOCD+strlen("XRevision: "));
-                       sprintf(buff, "%d", revision);
-                       version_str=buff;
+                       version_str = ZYLIN_OPENOCD;
                }
                else if (strcmp("zy1000", str) == 0)
                {
-                       version_str=ZYLIN_VERSION;
+                       version_str = ZYLIN_VERSION;
                }
                else if (strcmp("date", str) == 0)
                {
-                       version_str=ZYLIN_DATE;
+                       version_str = ZYLIN_DATE;
+               }
+               else if (strcmp("time", str) == 0)
+               {
+                       version_str = ZYLIN_TIME;
+               }
+               else if (strcmp("pcb", str) == 0)
+               {
+#ifdef CYGPKG_HAL_NIOS2
+                       version_str="c";
+#else
+                       version_str="b";
+#endif
+               }
+#ifdef CYGPKG_HAL_NIOS2
+               else if (strcmp("fpga", str) == 0)
+               {
+
+                       /* return a list of 32 bit integers to describe the expected
+                        * and actual FPGA
+                        */
+                       static char *fpga_id = "0x12345678 0x12345678 0x12345678 0x12345678";
+                       uint32_t id, timestamp;
+                       HAL_READ_UINT32(SYSID_BASE, id);
+                       HAL_READ_UINT32(SYSID_BASE+4, timestamp);
+                       sprintf(fpga_id, "0x%08x 0x%08x 0x%08x 0x%08x", id, timestamp, SYSID_ID, SYSID_TIMESTAMP);
+                       version_str = fpga_id;
+                       if ((argc>2) && (strcmp("time", str2) == 0))
+                       {
+                           time_t last_mod = timestamp;
+                           char * t = ctime (&last_mod) ;
+                           t[strlen(t)-1] = 0;
+                           version_str = t;
+                       }
                }
+#endif
+
                else
                {
                        return JIM_ERR;
@@ -293,58 +401,82 @@ static int jim_zy1000_version(Jim_Interp *interp, int argc, Jim_Obj *const *argv
 
        return JIM_OK;
 }
+#endif
 
+#ifdef CYGPKG_HAL_NIOS2
 
-static int
-zylinjtag_Jim_Command_powerstatus(Jim_Interp *interp,
-                                                                  int argc,
-               Jim_Obj * const *argv)
-{
-       if (argc != 1)
-       {
-               Jim_WrongNumArgs(interp, 1, argv, "powerstatus");
-               return JIM_ERR;
-       }
-
-       cyg_uint32 status;
-       ZY1000_PEEK(ZY1000_JTAG_BASE+0x10, status);
 
-       Jim_SetResult(interp, Jim_NewIntObj(interp, (status&0x80)!=0));
+struct info_forward
+{
+       void *data;
+       struct cyg_upgrade_info *upgraded_file;
+};
 
-       return JIM_OK;
+static void report_info(void *data, const char * format, va_list args)
+{
+       char *s = alloc_vprintf(format, args);
+       LOG_USER_N("%s", s);
+       free(s);
 }
 
-int zy1000_register_commands(struct command_context_s *cmd_ctx)
+struct cyg_upgrade_info firmware_info =
 {
-       register_command(cmd_ctx, NULL, "power", handle_power_command, COMMAND_ANY,
-                       "power <on/off> - turn power switch to target on/off. No arguments - print status.");
-
-       Jim_CreateCommand(interp, "zy1000_version", jim_zy1000_version, NULL, NULL);
-
+               (uint8_t *)0x84000000,
+               "/ram/firmware.phi",
+               "Firmware",
+               0x0300000,
+               0x1f00000 -
+               0x0300000,
+               "ZylinNiosFirmware\n",
+               report_info,
+};
 
-       Jim_CreateCommand(interp, "powerstatus", zylinjtag_Jim_Command_powerstatus, NULL, NULL);
+static int jim_zy1000_writefirmware(Jim_Interp *interp, int argc, Jim_Obj *const *argv)
+{
+       if (argc != 2)
+               return JIM_ERR;
 
-       return ERROR_OK;
-}
+       int length;
+       const char *str = Jim_GetString(argv[1], &length);
 
+       /* */
+       int tmpFile;
+       if ((tmpFile = open(firmware_info.file, O_RDWR | O_CREAT | O_TRUNC)) <= 0)
+       {
+               return JIM_ERR;
+       }
+       bool success;
+       success = write(tmpFile, str, length) == length;
+       close(tmpFile);
+       if (!success)
+               return JIM_ERR;
 
+       if (!cyg_firmware_upgrade(NULL, firmware_info))
+               return JIM_ERR;
 
+       return JIM_OK;
+}
+#endif
 
-int zy1000_init(void)
+static int
+zylinjtag_Jim_Command_powerstatus(Jim_Interp *interp,
+                                                                  int argc,
+               Jim_Obj * const *argv)
 {
-       LOG_USER("%s", ZYLIN_OPENOCD_VERSION);
+       if (argc != 1)
+       {
+               Jim_WrongNumArgs(interp, 1, argv, "powerstatus");
+               return JIM_ERR;
+       }
 
-       ZY1000_POKE(ZY1000_JTAG_BASE+0x10, 0x30); // Turn on LED1 & LED2
+       bool dropout = readPowerDropout();
 
-       setPower(true); // on by default
+       Jim_SetResult(interp, Jim_NewIntObj(interp, dropout));
 
+       return JIM_OK;
+}
 
-        /* deassert resets. Important to avoid infinite loop waiting for SRST to deassert */
-       zy1000_reset(0, 0);
-       zy1000_speed(jtag_get_speed());
 
-       return ERROR_OK;
-}
 
 int zy1000_quit(void)
 {
@@ -356,20 +488,35 @@ int zy1000_quit(void)
 
 int interface_jtag_execute_queue(void)
 {
-       cyg_uint32 empty;
+       uint32_t empty;
 
        waitIdle();
-       ZY1000_PEEK(ZY1000_JTAG_BASE+0x10, empty);
-       /* clear JTAG error register */
-       ZY1000_POKE(ZY1000_JTAG_BASE+0x14, 0x400);
 
-       if ((empty&0x400)!=0)
+       /* We must make sure to write data read back to memory location before we return
+        * from this fn
+        */
+       zy1000_flush_readqueue();
+
+       /* and handle any callbacks... */
+       zy1000_flush_callbackqueue();
+
+       if (zy1000_rclk)
        {
-               LOG_WARNING("RCLK timeout");
-               /* the error is informative only as we don't want to break the firmware if there
-                * is a false positive.
+               /* Only check for errors when using RCLK to speed up
+                * jtag over TCP/IP
                 */
-//             return ERROR_FAIL;
+               ZY1000_PEEK(ZY1000_JTAG_BASE + 0x10, empty);
+               /* clear JTAG error register */
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x14, 0x400);
+
+               if ((empty&0x400) != 0)
+               {
+                       LOG_WARNING("RCLK timeout");
+                       /* the error is informative only as we don't want to break the firmware if there
+                        * is a false positive.
+                        */
+       //              return ERROR_FAIL;
+               }
        }
        return ERROR_OK;
 }
@@ -377,167 +524,97 @@ int interface_jtag_execute_queue(void)
 
 
 
+static void writeShiftValue(uint8_t *data, int bits);
 
-static cyg_uint32 getShiftValue(void)
-{
-       cyg_uint32 value;
-       waitIdle();
-       ZY1000_PEEK(ZY1000_JTAG_BASE+0xc, value);
-       VERBOSE(LOG_INFO("getShiftValue %08x", value));
-       return value;
-}
-#if 0
-static cyg_uint32 getShiftValueFlip(void)
-{
-       cyg_uint32 value;
-       waitIdle();
-       ZY1000_PEEK(ZY1000_JTAG_BASE+0x18, value);
-       VERBOSE(LOG_INFO("getShiftValue %08x (flipped)", value));
-       return value;
-}
-#endif
-
-#if 0
-static void shiftValueInnerFlip(const tap_state_t state, const tap_state_t endState, int repeat, cyg_uint32 value)
-{
-       VERBOSE(LOG_INFO("shiftValueInner %s %s %d %08x (flipped)", tap_state_name(state), tap_state_name(endState), repeat, value));
-       cyg_uint32 a,b;
-       a=state;
-       b=endState;
-       ZY1000_POKE(ZY1000_JTAG_BASE+0xc, value);
-       ZY1000_POKE(ZY1000_JTAG_BASE+0x8, (1<<15)|(repeat<<8)|(a<<4)|b);
-       VERBOSE(getShiftValueFlip());
-}
-#endif
-
-extern int jtag_check_value(uint8_t *captured, void *priv);
-
-static __inline void scanFields(int num_fields, const scan_field_t *fields, tap_state_t shiftState, tap_state_t end_state)
+// here we shuffle N bits out/in
+static __inline void scanBits(const uint8_t *out_value, uint8_t *in_value, int num_bits, bool pause_now, tap_state_t shiftState, tap_state_t end_state)
 {
-       int i;
-       int j;
-       int k;
-
-       for (i = 0; i < num_fields; i++)
+       tap_state_t pause_state = shiftState;
+       for (int j = 0; j < num_bits; j += 32)
        {
-               cyg_uint32 value;
-
-               uint8_t *inBuffer=NULL;
-
-
-               // figure out where to store the input data
-               int num_bits=fields[i].num_bits;
-               if (fields[i].in_value!=NULL)
+               int k = num_bits - j;
+               if (k > 32)
                {
-                       inBuffer=fields[i].in_value;
+                       k = 32;
+                       /* we have more to shift out */
+               } else if (pause_now)
+               {
+                       /* this was the last to shift out this time */
+                       pause_state = end_state;
                }
 
-               // here we shuffle N bits out/in
-               j=0;
-               while (j<num_bits)
+               // we have (num_bits + 7)/8 bytes of bits to toggle out.
+               // bits are pushed out LSB to MSB
+               uint32_t value;
+               value = 0;
+               if (out_value != NULL)
                {
-                       tap_state_t pause_state;
-                       int l;
-                       k=num_bits-j;
-                       pause_state=(shiftState==TAP_DRSHIFT)?TAP_DRSHIFT:TAP_IRSHIFT;
-                       if (k>32)
-                       {
-                               k=32;
-                               /* we have more to shift out */
-                       } else if (i == num_fields-1)
+                       for (int l = 0; l < k; l += 8)
                        {
-                               /* this was the last to shift out this time */
-                               pause_state=end_state;
+                               value|=out_value[(j + l)/8]<<l;
                        }
+               }
+               /* mask away unused bits for easier debugging */
+               if (k < 32)
+               {
+                       value&=~(((uint32_t)0xffffffff) << k);
+               } else
+               {
+                       /* Shifting by >= 32 is not defined by the C standard
+                        * and will in fact shift by &0x1f bits on nios */
+               }
 
-                       // we have (num_bits+7)/8 bytes of bits to toggle out.
-                       // bits are pushed out LSB to MSB
-                       value=0;
-                       if (fields[i].out_value!=NULL)
-                       {
-                               for (l=0; l<k; l+=8)
-                               {
-                                       value|=fields[i].out_value[(j+l)/8]<<l;
-                               }
-                       }
-                       /* mask away unused bits for easier debugging */
-                       value&=~(((uint32_t)0xffffffff)<<k);
-
-                       shiftValueInner(shiftState, pause_state, k, value);
-
-                       if (inBuffer!=NULL)
-                       {
-                               // data in, LSB to MSB
-                               value=getShiftValue();
-                               // we're shifting in data to MSB, shift data to be aligned for returning the value
-                               value >>= 32-k;
+               shiftValueInner(shiftState, pause_state, k, value);
 
-                               for (l=0; l<k; l+=8)
-                               {
-                                       inBuffer[(j+l)/8]=(value>>l)&0xff;
-                               }
-                       }
-                       j+=k;
+               if (in_value != NULL)
+               {
+                       writeShiftValue(in_value + (j/8), k);
                }
        }
 }
 
-int interface_jtag_set_end_state(tap_state_t state)
+static __inline void scanFields(int num_fields, const struct scan_field *fields, tap_state_t shiftState, tap_state_t end_state)
 {
-       return ERROR_OK;
+       for (int i = 0; i < num_fields; i++)
+       {
+               scanBits(fields[i].out_value,
+                               fields[i].in_value,
+                               fields[i].num_bits,
+                               (i == num_fields-1),
+                               shiftState,
+                               end_state);
+       }
 }
 
-
-int interface_jtag_add_ir_scan(int num_fields, const scan_field_t *fields, tap_state_t state)
+int interface_jtag_add_ir_scan(struct jtag_tap *active, const struct scan_field *fields, tap_state_t state)
 {
-
-       int j;
        int scan_size = 0;
-       jtag_tap_t *tap, *nextTap;
-       for(tap = jtag_tap_next_enabled(NULL); tap!= NULL; tap=nextTap)
+       struct jtag_tap *tap, *nextTap;
+       tap_state_t pause_state = TAP_IRSHIFT;
+
+       for (tap = jtag_tap_next_enabled(NULL); tap!= NULL; tap = nextTap)
        {
-               nextTap=jtag_tap_next_enabled(tap);
-               tap_state_t end_state;
+               nextTap = jtag_tap_next_enabled(tap);
                if (nextTap==NULL)
                {
-                       end_state = state;
-               } else
-               {
-                       end_state = TAP_IRSHIFT;
+                       pause_state = state;
                }
-
-               int found = 0;
-
                scan_size = tap->ir_length;
 
                /* search the list */
-               for (j=0; j < num_fields; j++)
+               if (tap == active)
                {
-                       if (tap == fields[j].tap)
-                       {
-                               found = 1;
-
-                               scanFields(1, fields+j, TAP_IRSHIFT, end_state);
-                               /* update device information */
-                               buf_cpy(fields[j].out_value, tap->cur_instr, scan_size);
-
-                               tap->bypass = 0;
-                               break;
-                       }
-               }
+                       scanFields(1, fields, TAP_IRSHIFT, pause_state);
+                       /* update device information */
+                       buf_cpy(fields[0].out_value, tap->cur_instr, scan_size);
 
-               if (!found)
+                       tap->bypass = 0;
+               } else
                {
                        /* if a device isn't listed, set it to BYPASS */
-                       uint8_t ones[]={0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff};
+                       assert(scan_size <= 32);
+                       shiftValueInner(TAP_IRSHIFT, pause_state, scan_size, 0xffffffff);
 
-                       scan_field_t tmp;
-                       memset(&tmp, 0, sizeof(tmp));
-                       tmp.out_value = ones;
-                       tmp.num_bits = scan_size;
-                       scanFields(1, &tmp, TAP_IRSHIFT, end_state);
-                       /* update device information */
-                       buf_cpy(tmp.out_value, tap->cur_instr, scan_size);
                        tap->bypass = 1;
                }
        }
@@ -549,66 +626,46 @@ int interface_jtag_add_ir_scan(int num_fields, const scan_field_t *fields, tap_s
 
 
 
-int interface_jtag_add_plain_ir_scan(int num_fields, const scan_field_t *fields, tap_state_t state)
+int interface_jtag_add_plain_ir_scan(int num_bits, const uint8_t *out_bits, uint8_t *in_bits, tap_state_t state)
 {
-       scanFields(num_fields, fields, TAP_IRSHIFT, state);
-
+       scanBits(out_bits, in_bits, num_bits, true, TAP_IRSHIFT, state);
        return ERROR_OK;
 }
 
-/*extern jtag_command_t **jtag_get_last_command_p(void);*/
-
-int interface_jtag_add_dr_scan(int num_fields, const scan_field_t *fields, tap_state_t state)
+int interface_jtag_add_dr_scan(struct jtag_tap *active, int num_fields, const struct scan_field *fields, tap_state_t state)
 {
-
-       int j;
-       jtag_tap_t *tap, *nextTap;
-       for(tap = jtag_tap_next_enabled(NULL); tap!= NULL; tap=nextTap)
+       struct jtag_tap *tap, *nextTap;
+       tap_state_t pause_state = TAP_DRSHIFT;
+       for (tap = jtag_tap_next_enabled(NULL); tap!= NULL; tap = nextTap)
        {
-               nextTap=jtag_tap_next_enabled(tap);
-               int found=0;
-               tap_state_t end_state;
+               nextTap = jtag_tap_next_enabled(tap);
                if (nextTap==NULL)
                {
-                       end_state = state;
-               } else
-               {
-                       end_state = TAP_DRSHIFT;
+                       pause_state = state;
                }
 
-               for (j=0; j < num_fields; j++)
+               /* Find a range of fields to write to this tap */
+               if (tap == active)
                {
-                       if (tap == fields[j].tap)
-                       {
-                               found = 1;
-
-                               scanFields(1, fields+j, TAP_DRSHIFT, end_state);
-                       }
-               }
-               if (!found)
-               {
-                       scan_field_t tmp;
-                       /* program the scan field to 1 bit length, and ignore it's value */
-                       tmp.num_bits = 1;
-                       tmp.out_value = NULL;
-                       tmp.in_value = NULL;
+                       assert(!tap->bypass);
 
-                       scanFields(1, &tmp, TAP_DRSHIFT, end_state);
-               }
-               else
+                       scanFields(num_fields, fields, TAP_DRSHIFT, pause_state);
+               } else
                {
+                       /* Shift out a 0 for disabled tap's */
+                       assert(tap->bypass);
+                       shiftValueInner(TAP_DRSHIFT, pause_state, 1, 0);
                }
        }
        return ERROR_OK;
 }
 
-int interface_jtag_add_plain_dr_scan(int num_fields, const scan_field_t *fields, tap_state_t state)
+int interface_jtag_add_plain_dr_scan(int num_bits, const uint8_t *out_bits, uint8_t *in_bits, tap_state_t state)
 {
-       scanFields(num_fields, fields, TAP_DRSHIFT, state);
+       scanBits(out_bits, in_bits, num_bits, true, TAP_DRSHIFT, state);
        return ERROR_OK;
 }
 
-
 int interface_jtag_add_tlr()
 {
        setCurrentState(TAP_RESET);
@@ -616,11 +673,6 @@ int interface_jtag_add_tlr()
 }
 
 
-
-
-extern int jtag_nsrst_delay;
-extern int jtag_ntrst_delay;
-
 int interface_jtag_add_reset(int req_trst, int req_srst)
 {
        zy1000_reset(req_trst, req_srst);
@@ -634,13 +686,13 @@ static int zy1000_jtag_add_clocks(int num_cycles, tap_state_t state, tap_state_t
 
        /* execute num_cycles, 32 at the time. */
        int i;
-       for (i=0; i<num_cycles; i+=32)
+       for (i = 0; i < num_cycles; i += 32)
        {
                int num;
-               num=32;
-               if (num_cycles-i<num)
+               num = 32;
+               if (num_cycles-i < num)
                {
-                       num=num_cycles-i;
+                       num = num_cycles-i;
                }
                shiftValueInner(clockstate, clockstate, num, 0);
        }
@@ -649,7 +701,7 @@ static int zy1000_jtag_add_clocks(int num_cycles, tap_state_t state, tap_state_t
        /* finish in end_state */
        setCurrentState(state);
 #else
-       tap_state_t t=TAP_IDLE;
+       tap_state_t t = TAP_IDLE;
        /* test manual drive code on any target */
        int tms;
        uint8_t tms_scan = tap_get_tms_path(t, state);
@@ -659,13 +711,12 @@ static int zy1000_jtag_add_clocks(int num_cycles, tap_state_t state, tap_state_t
        {
                tms = (tms_scan >> i) & 1;
                waitIdle();
-               ZY1000_POKE(ZY1000_JTAG_BASE+0x28,  tms);
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x28,  tms);
        }
        waitIdle();
-       ZY1000_POKE(ZY1000_JTAG_BASE+0x20, state);
+       ZY1000_POKE(ZY1000_JTAG_BASE + 0x20, state);
 #endif
 
-
        return ERROR_OK;
 }
 
@@ -679,9 +730,36 @@ int interface_jtag_add_clocks(int num_cycles)
        return zy1000_jtag_add_clocks(num_cycles, cmd_queue_cur_state, cmd_queue_cur_state);
 }
 
-int interface_jtag_add_sleep(uint32_t us)
+int interface_add_tms_seq(unsigned num_bits, const uint8_t *seq, enum tap_state state)
 {
-       jtag_sleep(us);
+       /*wait for the fifo to be empty*/
+       waitIdle();
+
+       for (unsigned i = 0; i < num_bits; i++)
+       {
+               int tms;
+
+               if (((seq[i/8] >> (i % 8)) & 1) == 0)
+               {
+                       tms = 0;
+               }
+               else
+               {
+                       tms = 1;
+               }
+
+               waitIdle();
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, tms);
+       }
+
+       waitIdle();
+       if (state != TAP_INVALID)
+       {
+               ZY1000_POKE(ZY1000_JTAG_BASE + 0x20, state);
+       } else
+       {
+               /* this would be normal if we are switching to SWD mode */
+       }
        return ERROR_OK;
 }
 
@@ -690,12 +768,13 @@ int interface_jtag_add_pathmove(int num_states, const tap_state_t *path)
        int state_count;
        int tms = 0;
 
-       /*wait for the fifo to be empty*/
-       waitIdle();
-
        state_count = 0;
 
-       tap_state_t cur_state=cmd_queue_cur_state;
+       tap_state_t cur_state = cmd_queue_cur_state;
+
+       uint8_t seq[16];
+       memset(seq, 0, sizeof(seq));
+       assert(num_states < (int)((sizeof(seq) * 8)));
 
        while (num_states)
        {
@@ -713,57 +792,827 @@ int interface_jtag_add_pathmove(int num_states, const tap_state_t *path)
                        exit(-1);
                }
 
-               waitIdle();
-               ZY1000_POKE(ZY1000_JTAG_BASE+0x28,  tms);
+               seq[state_count/8] = seq[state_count/8] | (tms << (state_count % 8));
 
                cur_state = path[state_count];
                state_count++;
                num_states--;
        }
 
-       waitIdle();
-       ZY1000_POKE(ZY1000_JTAG_BASE+0x20,  cur_state);
-       return ERROR_OK;
+       return interface_add_tms_seq(state_count, seq, cur_state);
 }
 
-
-
-void embeddedice_write_dcc(jtag_tap_t *tap, int reg_addr, uint8_t *buffer, int little, int count)
+static void jtag_pre_post_bits(struct jtag_tap *tap, int *pre, int *post)
 {
-//     static int const reg_addr=0x5;
-       tap_state_t end_state=jtag_get_end_state();
-       if (jtag_tap_next_enabled(jtag_tap_next_enabled(NULL))==NULL)
+       /* bypass bits before and after */
+       int pre_bits = 0;
+       int post_bits = 0;
+
+       bool found = false;
+       struct jtag_tap *cur_tap, *nextTap;
+       for (cur_tap = jtag_tap_next_enabled(NULL); cur_tap!= NULL; cur_tap = nextTap)
        {
-               /* better performance via code duplication */
-               if (little)
+               nextTap = jtag_tap_next_enabled(cur_tap);
+               if (cur_tap == tap)
                {
-                       int i;
-                       for (i = 0; i < count; i++)
-                       {
-                               shiftValueInner(TAP_DRSHIFT, TAP_DRSHIFT, 32, fast_target_buffer_get_u32(buffer, 1));
-                               shiftValueInner(TAP_DRSHIFT, end_state, 6, reg_addr|(1<<5));
-                               buffer+=4;
-                       }
+                       found = true;
                } else
                {
-                       int i;
-                       for (i = 0; i < count; i++)
+                       if (found)
                        {
-                               shiftValueInner(TAP_DRSHIFT, TAP_DRSHIFT, 32, fast_target_buffer_get_u32(buffer, 0));
-                               shiftValueInner(TAP_DRSHIFT, end_state, 6, reg_addr|(1<<5));
-                               buffer+=4;
+                               post_bits++;
+                       } else
+                       {
+                               pre_bits++;
                        }
                }
        }
-       else
-       {
-               int i;
-               for (i = 0; i < count; i++)
-               {
-                       embeddedice_write_reg_inner(tap, reg_addr, fast_target_buffer_get_u32(buffer, little));
-                       buffer += 4;
+       *pre = pre_bits;
+       *post = post_bits;
+}
+
+/*
+       static const int embeddedice_num_bits[] = {32, 6};
+       uint32_t values[2];
+
+       values[0] = value;
+       values[1] = (1 << 5) | reg_addr;
+
+       jtag_add_dr_out(tap,
+                       2,
+                       embeddedice_num_bits,
+                       values,
+                       TAP_IDLE);
+*/
+
+void embeddedice_write_dcc(struct jtag_tap *tap, int reg_addr, uint8_t *buffer, int little, int count)
+{
+#if 0
+       int i;
+       for (i = 0; i < count; i++)
+       {
+               embeddedice_write_reg_inner(tap, reg_addr, fast_target_buffer_get_u32(buffer, little));
+               buffer += 4;
+       }
+#else
+       int pre_bits;
+       int post_bits;
+       jtag_pre_post_bits(tap, &pre_bits, &post_bits);
+
+       if ((pre_bits > 32) || (post_bits + 6 > 32))
+       {
+               int i;
+               for (i = 0; i < count; i++)
+               {
+                       embeddedice_write_reg_inner(tap, reg_addr, fast_target_buffer_get_u32(buffer, little));
+                       buffer += 4;
+               }
+       } else
+       {
+               int i;
+               for (i = 0; i < count; i++)
+               {
+                       /* Fewer pokes means we get to use the FIFO more efficiently */
+                       shiftValueInner(TAP_DRSHIFT, TAP_DRSHIFT, pre_bits, 0);
+                       shiftValueInner(TAP_DRSHIFT, TAP_DRSHIFT, 32, fast_target_buffer_get_u32(buffer, little));
+                       /* Danger! here we need to exit into the TAP_IDLE state to make
+                        * DCC pick up this value.
+                        */
+                       shiftValueInner(TAP_DRSHIFT, TAP_IDLE, 6 + post_bits, (reg_addr | (1 << 5)));
+                       buffer += 4;
+               }
+       }
+#endif
+}
+
+
+
+int arm11_run_instr_data_to_core_noack_inner(struct jtag_tap * tap, uint32_t opcode, uint32_t * data, size_t count)
+{
+       /* bypass bits before and after */
+       int pre_bits;
+       int post_bits;
+       jtag_pre_post_bits(tap, &pre_bits, &post_bits);
+       post_bits+=2;
+
+       if ((pre_bits > 32) || (post_bits > 32))
+       {
+               int arm11_run_instr_data_to_core_noack_inner_default(struct jtag_tap *, uint32_t, uint32_t *, size_t);
+               return arm11_run_instr_data_to_core_noack_inner_default(tap, opcode, data, count);
+       } else
+       {
+               static const int bits[] = {32, 2};
+               uint32_t values[] = {0, 0};
+
+               /* FIX!!!!!! the target_write_memory() API started this nasty problem
+                * with unaligned uint32_t * pointers... */
+               const uint8_t *t = (const uint8_t *)data;
+
+               while (--count > 0)
+               {
+#if 1
+                       /* Danger! This code doesn't update cmd_queue_cur_state, so
+                        * invoking jtag_add_pathmove() before jtag_add_dr_out() after
+                        * this loop would fail!
+                        */
+                       shiftValueInner(TAP_DRSHIFT, TAP_DRSHIFT, pre_bits, 0);
+
+                       uint32_t value;
+                       value = *t++;
+                       value |= (*t++<<8);
+                       value |= (*t++<<16);
+                       value |= (*t++<<24);
+
+                       shiftValueInner(TAP_DRSHIFT, TAP_DRSHIFT, 32, value);
+                       /* minimum 2 bits */
+                       shiftValueInner(TAP_DRSHIFT, TAP_DRPAUSE, post_bits, 0);
+
+                       /* copy & paste from arm11_dbgtap.c */
+                       //TAP_DREXIT2, TAP_DRUPDATE, TAP_IDLE, TAP_IDLE, TAP_IDLE, TAP_DRSELECT, TAP_DRCAPTURE, TAP_DRSHIFT
+                       /* KLUDGE! we have to flush the fifo or the Nios CPU locks up.
+                        * This is probably a bug in the Avalon bus(cross clocking bridge?)
+                        * or in the jtag registers module.
+                        */
+                       waitIdle();
+                       ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 1);
+                       ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 1);
+                       ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 0);
+                       ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 0);
+                       ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 0);
+                       ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 1);
+                       ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 0);
+                       ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 0);
+                       /* we don't have to wait for the queue to empty here */
+                       ZY1000_POKE(ZY1000_JTAG_BASE + 0x20, TAP_DRSHIFT);
+                       waitIdle();
+#else
+                       static const tap_state_t arm11_MOVE_DRPAUSE_IDLE_DRPAUSE_with_delay[] =
+                       {
+                               TAP_DREXIT2, TAP_DRUPDATE, TAP_IDLE, TAP_IDLE, TAP_IDLE, TAP_DRSELECT, TAP_DRCAPTURE, TAP_DRSHIFT
+                       };
+
+                       values[0] = *t++;
+                       values[0] |= (*t++<<8);
+                       values[0] |= (*t++<<16);
+                       values[0] |= (*t++<<24);
+
+                       jtag_add_dr_out(tap,
+                               2,
+                               bits,
+                               values,
+                               TAP_IDLE);
+
+                       jtag_add_pathmove(ARRAY_SIZE(arm11_MOVE_DRPAUSE_IDLE_DRPAUSE_with_delay),
+                               arm11_MOVE_DRPAUSE_IDLE_DRPAUSE_with_delay);
+#endif
+               }
+
+               values[0] = *t++;
+               values[0] |= (*t++<<8);
+               values[0] |= (*t++<<16);
+               values[0] |= (*t++<<24);
+
+               /* This will happen on the last iteration updating cmd_queue_cur_state
+                * so we don't have to track it during the common code path
+                */
+               jtag_add_dr_out(tap,
+                       2,
+                       bits,
+                       values,
+                       TAP_IDLE);
+
+               return jtag_execute_queue();
+       }
+}
+
+
+static const struct command_registration zy1000_commands[] = {
+       {
+               .name = "power",
+               .handler = handle_power_command,
+               .mode = COMMAND_ANY,
+               .help = "Turn power switch to target on/off. "
+                       "With no arguments, prints status.",
+               .usage = "('on'|'off)",
+       },
+#if BUILD_ECOSBOARD
+       {
+               .name = "zy1000_version",
+               .mode = COMMAND_ANY,
+               .jim_handler = jim_zy1000_version,
+               .help = "Print version info for zy1000.",
+               .usage = "['openocd'|'zy1000'|'date'|'time'|'pcb'|'fpga']",
+       },
+#else
+       {
+               .name = "zy1000_server",
+               .mode = COMMAND_ANY,
+               .jim_handler = jim_zy1000_server,
+               .help = "Tcpip address for ZY1000 server.",
+               .usage = "address",
+       },
+#endif
+       {
+               .name = "powerstatus",
+               .mode = COMMAND_ANY,
+               .jim_handler = zylinjtag_Jim_Command_powerstatus,
+               .help = "Returns power status of target",
+       },
+#ifdef CYGPKG_HAL_NIOS2
+       {
+               .name = "updatezy1000firmware",
+               .mode = COMMAND_ANY,
+               .jim_handler = jim_zy1000_writefirmware,
+               .help = "writes firmware to flash",
+               /* .usage = "some_string", */
+       },
+#endif
+       COMMAND_REGISTRATION_DONE
+};
+
+
+static int tcp_ip = -1;
+
+/* Write large packets if we can */
+static size_t out_pos;
+static uint8_t out_buffer[16384];
+static size_t in_pos;
+static size_t in_write;
+static uint8_t in_buffer[16384];
+
+static bool flush_writes(void)
+{
+       bool ok = (write(tcp_ip, out_buffer, out_pos) == (int)out_pos);
+       out_pos = 0;
+       return ok;
+}
+
+static bool writeLong(uint32_t l)
+{
+       int i;
+       for (i = 0; i < 4; i++)
+       {
+               uint8_t c = (l >> (i*8))&0xff;
+               out_buffer[out_pos++] = c;
+               if (out_pos >= sizeof(out_buffer))
+               {
+                       if (!flush_writes())
+                       {
+                               return false;
+                       }
+               }
+       }
+       return true;
+}
+
+static bool readLong(uint32_t *out_data)
+{
+       if (out_pos > 0)
+       {
+               if (!flush_writes())
+               {
+                       return false;
+               }
+       }
+
+       uint32_t data = 0;
+       int i;
+       for (i = 0; i < 4; i++)
+       {
+               uint8_t c;
+               if (in_pos == in_write)
+               {
+                       /* read more */
+                       int t;
+                       t = read(tcp_ip, in_buffer, sizeof(in_buffer));
+                       if (t < 1)
+                       {
+                               return false;
+                       }
+                       in_write = (size_t) t;
+                       in_pos = 0;
+               }
+               c = in_buffer[in_pos++];
+
+               data |= (c << (i*8));
+       }
+       *out_data = data;
+       return true;
+}
+
+enum ZY1000_CMD
+{
+       ZY1000_CMD_POKE = 0x0,
+       ZY1000_CMD_PEEK = 0x8,
+       ZY1000_CMD_SLEEP = 0x1,
+       ZY1000_CMD_WAITIDLE = 2
+};
+
+
+#if !BUILD_ECOSBOARD
+
+#include <sys/socket.h> /* for socket(), connect(), send(), and recv() */
+#include <arpa/inet.h>  /* for sockaddr_in and inet_addr() */
+
+/* We initialize this late since we need to know the server address
+ * first.
+ */
+static void tcpip_open(void)
+{
+       if (tcp_ip >= 0)
+               return;
+
+       struct sockaddr_in echoServAddr; /* Echo server address */
+
+       /* Create a reliable, stream socket using TCP */
+       if ((tcp_ip = socket(PF_INET, SOCK_STREAM, IPPROTO_TCP)) < 0)
+       {
+               fprintf(stderr, "Failed to connect to zy1000 server\n");
+               exit(-1);
+       }
+
+       /* Construct the server address structure */
+       memset(&echoServAddr, 0, sizeof(echoServAddr)); /* Zero out structure */
+       echoServAddr.sin_family = AF_INET; /* Internet address family */
+       echoServAddr.sin_addr.s_addr = inet_addr(tcp_server); /* Server IP address */
+       echoServAddr.sin_port = htons(7777); /* Server port */
+
+       /* Establish the connection to the echo server */
+       if (connect(tcp_ip, (struct sockaddr *) &echoServAddr, sizeof(echoServAddr)) < 0)
+       {
+               fprintf(stderr, "Failed to connect to zy1000 server\n");
+               exit(-1);
+       }
+
+       int flag = 1;
+       setsockopt(tcp_ip,      /* socket affected */
+                       IPPROTO_TCP,            /* set option at TCP level */
+                       TCP_NODELAY,            /* name of option */
+                       (char *)&flag,          /* the cast is historical cruft */
+                       sizeof(int));           /* length of option value */
+
+}
+
+
+/* send a poke */
+void zy1000_tcpout(uint32_t address, uint32_t data)
+{
+       tcpip_open();
+       if (!writeLong((ZY1000_CMD_POKE << 24) | address)||
+                       !writeLong(data))
+       {
+               fprintf(stderr, "Could not write to zy1000 server\n");
+               exit(-1);
+       }
+}
+
+/* By sending the wait to the server, we avoid a readback
+ * of status. Radically improves performance for this operation
+ * with long ping times.
+ */
+void waitIdle(void)
+{
+       tcpip_open();
+       if (!writeLong((ZY1000_CMD_WAITIDLE << 24)))
+       {
+               fprintf(stderr, "Could not write to zy1000 server\n");
+               exit(-1);
+       }
+}
+
+uint32_t zy1000_tcpin(uint32_t address)
+{
+       tcpip_open();
+
+       zy1000_flush_readqueue();
+
+       uint32_t data;
+       if (!writeLong((ZY1000_CMD_PEEK << 24) | address)||
+                       !readLong(&data))
+       {
+               fprintf(stderr, "Could not read from zy1000 server\n");
+               exit(-1);
+       }
+       return data;
+}
+
+int interface_jtag_add_sleep(uint32_t us)
+{
+       tcpip_open();
+       if (!writeLong((ZY1000_CMD_SLEEP << 24))||
+                       !writeLong(us))
+       {
+               fprintf(stderr, "Could not read from zy1000 server\n");
+               exit(-1);
+       }
+       return ERROR_OK;
+}
+
+/* queue a readback */
+#define readqueue_size 16384
+static struct
+{
+       uint8_t *dest;
+       int bits;
+} readqueue[readqueue_size];
+
+static int readqueue_pos = 0;
+
+/* flush the readqueue, this means reading any data that
+ * we're expecting and store them into the final position
+ */
+void zy1000_flush_readqueue(void)
+{
+       if (readqueue_pos == 0)
+       {
+               /* simply debugging by allowing easy breakpoints when there
+                * is something to do. */
+               return;
+       }
+       int i;
+       tcpip_open();
+       for (i = 0; i < readqueue_pos; i++)
+       {
+               uint32_t value;
+               if (!readLong(&value))
+               {
+                       fprintf(stderr, "Could not read from zy1000 server\n");
+                       exit(-1);
+               }
+
+               uint8_t *in_value = readqueue[i].dest;
+               int k = readqueue[i].bits;
+
+               // we're shifting in data to MSB, shift data to be aligned for returning the value
+               value >>= 32-k;
+
+               for (int l = 0; l < k; l += 8)
+               {
+                       in_value[l/8]=(value >> l)&0xff;
                }
        }
+       readqueue_pos = 0;
 }
 
+/* By queuing the callback's we avoid flushing the
+read queue until jtag_execute_queue(). This can
+reduce latency dramatically for cases where
+callbacks are used extensively.
+*/
+#define callbackqueue_size 128
+static struct callbackentry
+{
+       jtag_callback_t callback;
+       jtag_callback_data_t data0;
+       jtag_callback_data_t data1;
+       jtag_callback_data_t data2;
+       jtag_callback_data_t data3;
+} callbackqueue[callbackqueue_size];
+
+static int callbackqueue_pos = 0;
+
+void zy1000_jtag_add_callback4(jtag_callback_t callback, jtag_callback_data_t data0, jtag_callback_data_t data1, jtag_callback_data_t data2, jtag_callback_data_t data3)
+{
+       if (callbackqueue_pos >= callbackqueue_size)
+       {
+               zy1000_flush_callbackqueue();
+       }
+
+       callbackqueue[callbackqueue_pos].callback = callback;
+       callbackqueue[callbackqueue_pos].data0 = data0;
+       callbackqueue[callbackqueue_pos].data1 = data1;
+       callbackqueue[callbackqueue_pos].data2 = data2;
+       callbackqueue[callbackqueue_pos].data3 = data3;
+       callbackqueue_pos++;
+}
 
+static int zy1000_jtag_convert_to_callback4(jtag_callback_data_t data0, jtag_callback_data_t data1, jtag_callback_data_t data2, jtag_callback_data_t data3)
+{
+       ((jtag_callback1_t)data1)(data0);
+       return ERROR_OK;
+}
+
+void zy1000_jtag_add_callback(jtag_callback1_t callback, jtag_callback_data_t data0)
+{
+       zy1000_jtag_add_callback4(zy1000_jtag_convert_to_callback4, data0, (jtag_callback_data_t)callback, 0, 0);
+}
+
+void zy1000_flush_callbackqueue(void)
+{
+       /* we have to flush the read queue so we have access to
+        the data the callbacks will use 
+       */
+       zy1000_flush_readqueue();
+       int i;
+       for (i = 0; i < callbackqueue_pos; i++)
+       {
+               struct callbackentry *entry = &callbackqueue[i];
+               jtag_set_error(entry->callback(entry->data0, entry->data1, entry->data2, entry->data3));
+       }
+       callbackqueue_pos = 0;
+}
+
+static void writeShiftValue(uint8_t *data, int bits)
+{
+       waitIdle();
+
+       if (!writeLong((ZY1000_CMD_PEEK << 24) | (ZY1000_JTAG_BASE + 0xc)))
+       {
+               fprintf(stderr, "Could not read from zy1000 server\n");
+               exit(-1);
+       }
+
+       if (readqueue_pos >= readqueue_size)
+       {
+               zy1000_flush_readqueue();
+       }
+
+       readqueue[readqueue_pos].dest = data;
+       readqueue[readqueue_pos].bits = bits;
+       readqueue_pos++;
+}
+
+#else
+
+static void writeShiftValue(uint8_t *data, int bits)
+{
+       uint32_t value;
+       waitIdle();
+       ZY1000_PEEK(ZY1000_JTAG_BASE + 0xc, value);
+       VERBOSE(LOG_INFO("getShiftValue %08x", value));
+
+       // data in, LSB to MSB
+       // we're shifting in data to MSB, shift data to be aligned for returning the value
+       value >>= 32 - bits;
+
+       for (int l = 0; l < bits; l += 8)
+       {
+               data[l/8]=(value >> l)&0xff;
+       }
+}
+
+#endif
+
+#if BUILD_ECOSBOARD
+static char tcpip_stack[2048];
+static cyg_thread tcpip_thread_object;
+static cyg_handle_t tcpip_thread_handle;
+
+static char watchdog_stack[2048];
+static cyg_thread watchdog_thread_object;
+static cyg_handle_t watchdog_thread_handle;
+
+/* Infinite loop peeking & poking */
+static void tcpipserver(void)
+{
+       for (;;)
+       {
+               uint32_t address;
+               if (!readLong(&address))
+                       return;
+               enum ZY1000_CMD c = (address >> 24) & 0xff;
+               address &= 0xffffff;
+               switch (c)
+               {
+                       case ZY1000_CMD_POKE:
+                       {
+                               uint32_t data;
+                               if (!readLong(&data))
+                                       return;
+                               address &= ~0x80000000;
+                               ZY1000_POKE(address + ZY1000_JTAG_BASE, data);
+                               break;
+                       }
+                       case ZY1000_CMD_PEEK:
+                       {
+                               uint32_t data;
+                               ZY1000_PEEK(address + ZY1000_JTAG_BASE, data);
+                               if (!writeLong(data))
+                                       return;
+                               break;
+                       }
+                       case ZY1000_CMD_SLEEP:
+                       {
+                               uint32_t data;
+                               if (!readLong(&data))
+                                       return;
+                               jtag_sleep(data);
+                               break;
+                       }
+                       case ZY1000_CMD_WAITIDLE:
+                       {
+                               waitIdle();
+                               break;
+                       }
+                       default:
+                               return;
+               }
+       }
+}
+
+
+static void tcpip_server(cyg_addrword_t data)
+{
+       int so_reuseaddr_option = 1;
+
+       int fd;
+       if ((fd = socket(AF_INET, SOCK_STREAM, 0)) == -1)
+       {
+               LOG_ERROR("error creating socket: %s", strerror(errno));
+               exit(-1);
+       }
+
+       setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (void*) &so_reuseaddr_option,
+                       sizeof(int));
+
+       struct sockaddr_in sin;
+       unsigned int address_size;
+       address_size = sizeof(sin);
+       memset(&sin, 0, sizeof(sin));
+       sin.sin_family = AF_INET;
+       sin.sin_addr.s_addr = INADDR_ANY;
+       sin.sin_port = htons(7777);
+
+       if (bind(fd, (struct sockaddr *) &sin, sizeof(sin)) == -1)
+       {
+               LOG_ERROR("couldn't bind to socket: %s", strerror(errno));
+               exit(-1);
+       }
+
+       if (listen(fd, 1) == -1)
+       {
+               LOG_ERROR("couldn't listen on socket: %s", strerror(errno));
+               exit(-1);
+       }
+
+
+       for (;;)
+       {
+               tcp_ip = accept(fd, (struct sockaddr *) &sin, &address_size);
+               if (tcp_ip < 0)
+               {
+                       continue;
+               }
+
+               int flag = 1;
+               setsockopt(tcp_ip,      /* socket affected */
+                               IPPROTO_TCP,            /* set option at TCP level */
+                               TCP_NODELAY,            /* name of option */
+                               (char *)&flag,          /* the cast is historical cruft */
+                               sizeof(int));           /* length of option value */
+
+               bool save_poll = jtag_poll_get_enabled();
+
+               /* polling will screw up the "connection" */
+               jtag_poll_set_enabled(false);
+
+               tcpipserver();
+
+               jtag_poll_set_enabled(save_poll);
+
+               close(tcp_ip);
+
+       }
+       close(fd);
+
+}
+
+#ifdef WATCHDOG_BASE
+/* If we connect to port 8888 we must send a char every 10s or the board resets itself */
+static void watchdog_server(cyg_addrword_t data)
+{
+       int so_reuseaddr_option = 1;
+
+       int fd;
+       if ((fd = socket(AF_INET, SOCK_STREAM, 0)) == -1)
+       {
+               LOG_ERROR("error creating socket: %s", strerror(errno));
+               exit(-1);
+       }
+
+       setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (void*) &so_reuseaddr_option,
+                       sizeof(int));
+
+       struct sockaddr_in sin;
+       unsigned int address_size;
+       address_size = sizeof(sin);
+       memset(&sin, 0, sizeof(sin));
+       sin.sin_family = AF_INET;
+       sin.sin_addr.s_addr = INADDR_ANY;
+       sin.sin_port = htons(8888);
+
+       if (bind(fd, (struct sockaddr *) &sin, sizeof(sin)) == -1)
+       {
+               LOG_ERROR("couldn't bind to socket: %s", strerror(errno));
+               exit(-1);
+       }
+
+       if (listen(fd, 1) == -1)
+       {
+               LOG_ERROR("couldn't listen on socket: %s", strerror(errno));
+               exit(-1);
+       }
+
+
+       for (;;)
+       {
+               int watchdog_ip = accept(fd, (struct sockaddr *) &sin, &address_size);
+
+               /* Start watchdog, must be reset every 10 seconds. */
+               HAL_WRITE_UINT32(WATCHDOG_BASE + 4, 4);
+
+               if (watchdog_ip < 0)
+               {
+                       LOG_ERROR("couldn't open watchdog socket: %s", strerror(errno));
+                       exit(-1);
+               }
+
+               int flag = 1;
+               setsockopt(watchdog_ip, /* socket affected */
+                               IPPROTO_TCP,            /* set option at TCP level */
+                               TCP_NODELAY,            /* name of option */
+                               (char *)&flag,          /* the cast is historical cruft */
+                               sizeof(int));           /* length of option value */
+
+
+               char buf;
+               for (;;)
+               {
+                       if (read(watchdog_ip, &buf, 1) == 1)
+                       {
+                               /* Reset timer */
+                               HAL_WRITE_UINT32(WATCHDOG_BASE + 8, 0x1234);
+                               /* Echo so we can telnet in and see that resetting works */
+                               write(watchdog_ip, &buf, 1);
+                       } else
+                       {
+                               /* Stop tickling the watchdog, the CPU will reset in < 10 seconds
+                                * now.
+                                */
+                               return;
+                       }
+
+               }
+
+               /* Never reached */
+       }
+}
+#endif
+
+int interface_jtag_add_sleep(uint32_t us)
+{
+       jtag_sleep(us);
+       return ERROR_OK;
+}
+
+#endif
+
+
+int zy1000_init(void)
+{
+#if BUILD_ECOSBOARD
+       LOG_USER("%s", ZYLIN_OPENOCD_VERSION);
+#endif
+
+       ZY1000_POKE(ZY1000_JTAG_BASE + 0x10, 0x30); // Turn on LED1 & LED2
+
+       setPower(true); // on by default
+
+
+        /* deassert resets. Important to avoid infinite loop waiting for SRST to deassert */
+       zy1000_reset(0, 0);
+       zy1000_speed(jtag_get_speed());
+
+
+#if BUILD_ECOSBOARD
+       cyg_thread_create(1, tcpip_server, (cyg_addrword_t) 0, "tcip/ip server",
+                       (void *) tcpip_stack, sizeof(tcpip_stack),
+                       &tcpip_thread_handle, &tcpip_thread_object);
+       cyg_thread_resume(tcpip_thread_handle);
+#ifdef WATCHDOG_BASE
+       cyg_thread_create(1, watchdog_server, (cyg_addrword_t) 0, "watchdog tcip/ip server",
+                       (void *) watchdog_stack, sizeof(watchdog_stack),
+                       &watchdog_thread_handle, &watchdog_thread_object);
+       cyg_thread_resume(watchdog_thread_handle);
+#endif
+#endif
+
+       return ERROR_OK;
+}
+
+
+
+struct jtag_interface zy1000_interface =
+{
+       .name = "ZY1000",
+       .supported = DEBUG_CAP_TMS_SEQ,
+       .execute_queue = NULL,
+       .speed = zy1000_speed,
+       .commands = zy1000_commands,
+       .init = zy1000_init,
+       .quit = zy1000_quit,
+       .khz = zy1000_khz,
+       .speed_div = zy1000_speed_div,
+       .power_dropout = zy1000_power_dropout,
+       .srst_asserted = zy1000_srst_asserted,
+};