Merge branch 'config_core_selector'

1 files changed, 543 insertions, 0 deletions

diff --git a/config/config.py b/config/config.py
new file mode 100755
index 0000000..0f3818f
--- /dev/null
+++ b/config/config.py
@@ -0,0 +1,543 @@
+#!/usr/bin/env python
+
+"""
+Generate core_selector.v and core_vfiles.mk for a set of cores.
+"""
+
+# History of Cryptech bus addressing scheme, as best I understand it.
+#
+# The old old addressing scheme that Joachim and Paul came up with
+# was:
+#
+#  3 bits of segment  selector [16:14]
+#  6 bits of core     selector [13:8]
+#  8 bits of register selector [7:0]
+#
+# modexp6s needed more register bits than that, so Pavel changed
+# addressing within the math segment to:
+#
+#  3 bits of segment  selector [16:14]
+#  4 bits of core     selector [13:10]
+# 10 bits of register selector [9:0]
+#
+# Meanwhile, Paul eliminated segments entirely when writing the
+# ancestor of this script, resulting in:
+#
+#  9 bits of core     selector [16:8]
+#  8 bits of register selector [7:0]
+#
+# Taking Pavel's and Paul's changes together, we'd get:
+#
+#  7 bits of core     selector [16:10]
+# 10 bits of register selector [9:0]
+#
+# Except that this would waste space for most cores, and make things
+# very confusing for the TRNG cores.  So, instead, we keep Paul's
+# two-level (no segment) scheme and handle modexps6 as a set of four
+# contiguous "cores" with a 10-bit composite register selector.
+
+# At present, TRNG core's internal multiplexer doesn't allocate cores
+# contiguously, there's a gap, and one just has to know what the
+# offsets are.  Current theory is that we'll fix the TRNG core to get
+# rid of this problem, but for now the workaround requires this script
+# to know the magic offsets for the high 4 bits of the 12-bit TRNG
+# address:
+#
+# 0x0: trng
+# 0x5: entropy1 (avalanche)
+# 0x6: entropy2 (rosc)
+# 0xa: mixer
+# 0xb: csprng
+
+# The modexps6 core drags in a one clock cycle delay to other cores,
+# to compensate for the extra clock cycle consumed by the block
+# memories used in the modexps6 core.  We probably want a general
+# solution for this, because we're going to run into this problem for
+# any core that handles arguments big enough to require block memory.
+
+# To Do:
+#
+# - Consider automating the one-clock-cycle delay stuff by adding
+#   another boolean flag to the config file.  Default would be no
+#   delay, if any included core sets the "I use block memories" flag,
+#   all other cores would get the delay.  Slightly tedious but
+#   something we can calculate easily enough, and probably an
+#   improvement over wiring in the delay when nothing needs it.
+#
+# - Rename script and its config file to something more meaningful.
+#
+# - Figure out whether this really belongs in the novena repository at
+#   all, seems more generic than that.
+
+
+def main():
+    """
+    Parse arguments and config file, generate core list, generate output.
+    """
+
+    from argparse import ArgumentParser, FileType, ArgumentDefaultsHelpFormatter
+    from sys      import exit
+
+    parser = ArgumentParser(description = __doc__, formatter_class = ArgumentDefaultsHelpFormatter)
+    parser.add_argument("-d", "--debug",   help = "enable debugging",   action = "store_true")
+    parser.add_argument("-s", "--section", help = "config file section")
+    parser.add_argument("-c", "--config",  help = "configuration file",   default = "config.cfg",       type = FileType("r"))
+    parser.add_argument("--verilog",       help = "verilog output file",  default = "core_selector.v",  type = FileType("w"))
+    parser.add_argument("--makefile",      help = "output makefile",      default = "core_vfiles.mk",   type = FileType("w"))
+    parser.add_argument("core",            help = "name(s) of core(s)", nargs = "*")
+    args = parser.parse_args()
+
+    try:
+        cfg = RawConfigParser()
+        cfg.readfp(args.config)
+
+        if args.core:
+            cores = args.core
+        else:
+            section = args.section or cfg.get("default", "default-section")
+            cores = cfg.get(section, "cores").split()
+
+        cores.insert(0, "board_regs")
+        cores.insert(1, "comm_regs")
+
+        cores = tuple(Core.new(core) for core in cores)
+
+        core_number = 0
+        for core in cores:
+            core_number = core.assign_core_number(core_number)
+
+        cores[0].reset_high = True
+        cores[1].reset_high = True
+
+        for core in cores:
+            core.configure(cfg)
+
+        if False:
+
+            # For some reason, attempting to optimize out the delay
+            # code entirely results in a non-working bitstream.  Don't
+            # know why, disabling the optimization works, so just do
+            # that for now.
+
+            Core.need_one_cycle_delay = any(core.block_memory for core in cores)
+
+        args.verilog.write(createModule_template.format(
+            addrs = "".join(core.createAddr()     for core in cores),
+            insts = "".join(core.createInstance() for core in cores),
+            muxes = "".join(core.createMux()      for core in cores)))
+
+        args.makefile.write(listVfiles_template.format(
+            vfiles = "".join(core.listVfiles()    for core in cores)))
+
+    except Exception, e:
+        if args.debug:
+            raise
+        exit(str(e))
+
+
+try:
+    import ConfigParser as configparser
+except ImportError:
+    import configparser   
+
+class RawConfigParser(configparser.RawConfigParser):
+    """
+    RawConfigParser with a few extensions.
+    """
+
+    def getboolean(self, section, option, default = False):
+        if self.has_option(section, option):
+            # RawConfigParser is an old-stle class, super() doesn't work, feh.
+            return configparser.RawConfigParser.getboolean(self, section, option)
+        else:
+            return default
+
+    def getvalues(self, section, option):
+        if self.has_option(section, option):
+            for value in self.get(section, option).split():
+                yield value
+
+
+class Core(object):
+    """
+    Data and methods for a generic core.  We can use this directly for
+    most cores, a few are weird and require subclassing to override
+    particular methods.
+    """
+
+    # Class variable tracking how many times a particular core has
+    # been instantiated.  This controls instance numbering.
+
+    _instance_count = {}
+
+    # Class variable mapping core name to subclass for special cases.
+
+    special_class = {}
+
+    # Class variable recording whether we need a one-cycle delay to
+    # compensate for block memories.
+
+    need_one_cycle_delay = True
+
+    def __init__(self, name):
+        self.name = name
+        self.core_number = None
+        self.vfiles = []
+        self.reset_high = False
+        self.error_wire = True
+        self.block_memory = False
+        self.instance_number = self._instance_count.get(name, 0)
+        self._instance_count[name] = self.instance_number + 1
+
+    @classmethod
+    def new(cls, name):
+        return cls.special_class.get(name, cls)(name)
+
+    def assign_core_number(self, n):
+        self.core_number = n
+        return n + 1
+
+    def configure(self, cfg):
+        if self.instance_number == 0:
+            self.vfiles.extend(cfg.getvalues(self.name, "vfiles"))
+            for required in cfg.getvalues(self.name, "requires"):
+                if required not in self._instance_count:
+                    self.vfiles.extend(cfg.getvalues(required, "vfiles"))
+        self.reset_high = cfg.getboolean(self.name, "reset_high", self.reset_high)
+        self.error_wire = cfg.getboolean(self.name, "error_wire", self.error_wire)
+        self.block_memory = cfg.getboolean(self.name, "block_memory", self.block_memory)
+
+    @property
+    def instance_name(self):
+        if self._instance_count[self.name] > 1:
+            return "{}_{}".format(self.name, self.instance_number)
+        else:
+            return self.name
+
+    @property
+    def upper_instance_name(self):
+        return self.instance_name.upper()
+
+    @property
+    def reset_pin(self):
+        return ".rst(sys_rst)" if self.reset_high else ".reset_n(~sys_rst)"
+
+    @property
+    def error_port(self):
+        return ",\n      .error(error_{core.instance_name})".format(core = self) if self.error_wire else ""
+
+    @property
+    def one_cycle_delay(self):
+        return one_cycle_delay_template.format(core = self) if self.need_one_cycle_delay and not self.block_memory else ""
+
+    @property
+    def mux_data_reg(self):
+        return "read_data_" + self.instance_name + ("_reg" if self.need_one_cycle_delay and not self.block_memory else "")
+
+    @property
+    def mux_error_reg(self):
+        return "error_" + self.instance_name if self.error_wire else "0"
+
+    def createInstance(self):
+        return createInstance_template_generic.format(core = self)
+
+    def createAddr(self):
+        return createAddr_template.format(core = self)
+
+    def createMux(self):
+        return createMux_template.format(core = self, core0 = self)
+
+    def listVfiles(self):
+        return "".join(" \\\n\t$(CORE_TREE)/" + vfile for vfile in self.vfiles)
+
+
+class SubCore(Core):
+    """"
+    Override mux handling for TRNG's sub-cores.
+    """
+
+    def __init__(self, name, parent):
+        super(SubCore, self).__init__(name)
+        self.parent = parent
+
+    def createMux(self):
+        return createMux_template.format(core = self, core0 = self.parent)
+
+
+class TRNGCore(Core):
+    """
+    The TRNG core has an internal mux with slots for 15 sub-cores,
+    most of which are empty.  This is a bit of a mess.
+
+    Mostly this means that our method calls have to iterate over all
+    of the subcores after handling the base TRNG core, but we also use
+    different templates, and fiddle with addresses a bit.
+
+    Mux numbers have to be dug out of the TRNG Verilog source.
+    """
+
+    # TRNG subcore name -> internal mux number.
+    subcore_parameters = dict(avalanche_entropy = 0x5,
+                              rosc_entropy      = 0x6,
+                              trng_mixer        = 0xa,
+                              trng_csprng       = 0xb)
+
+    def __init__(self, name):
+        super(TRNGCore, self).__init__(name)
+        self.subcores = tuple(SubCore(name, self)
+                              for name in sorted(self.subcore_parameters,
+                                                 key = lambda x: self.subcore_parameters[x]))
+
+    def assign_core_number(self, n):
+        n = super(TRNGCore, self).assign_core_number(n)
+        for subcore in self.subcores:
+            subcore.assign_core_number(self.core_number + self.subcore_parameters[subcore.name])
+        return n + 15
+
+    @property
+    def last_subcore_upper_instance_name(self):
+        return self.subcores[-1].upper_instance_name
+
+    def createInstance(self):
+        return createInstance_template_TRNG.format(core = self)
+
+    def createAddr(self):
+        return super(TRNGCore, self).createAddr() + "".join(subcore.createAddr() for subcore in self.subcores)
+
+    def createMux(self):
+        return super(TRNGCore, self).createMux() + "".join(subcore.createMux() for subcore in self.subcores)
+
+
+class ModExpS6Core(Core):
+    """
+    ModExpS6 core consumes as much space as four ordinary cores, and
+    uses different templates to handle the differences in timing and
+    addressing.
+    """
+
+    def assign_core_number(self, n):
+        n = super(ModExpS6Core, self).assign_core_number(n)
+        return n + 3
+
+    def createInstance(self):
+        return createInstance_template_ModExpS6.format(core = self)
+
+    def createMux(self):
+        return createMux_modexps6_template.format(core = self, core0 = self)
+
+
+# Hook special classes in as handlers for the cores that require them.
+
+Core.special_class.update(
+    trng        = TRNGCore,
+    modexps6    = ModExpS6Core)
+
+
+# Templates (format strings), here instead of inline in the functions
+# that use them, both because some of these are shared between
+# multiple functions and because it's easier to read these (and get
+# the indentation right) when the're separate.
+
+# Template used by .createAddr() methods.
+
+createAddr_template = """\
+   localparam   CORE_ADDR_{core.upper_instance_name:21s} = 9'h{core.core_number:02x};
+"""
+
+# Template used by Core.createInstance().
+
+createInstance_template_generic = """\
+   //----------------------------------------------------------------
+   // {core.upper_instance_name}
+   //----------------------------------------------------------------
+   wire                 enable_{core.instance_name} = (addr_core_num == CORE_ADDR_{core.upper_instance_name});
+   wire [31: 0]         read_data_{core.instance_name};
+   wire                 error_{core.instance_name};
+
+   {core.name} {core.instance_name}_inst
+     (
+      .clk(sys_clk),
+      {core.reset_pin},
+
+      .cs(enable_{core.instance_name} & (sys_eim_rd | sys_eim_wr)),
+      .we(sys_eim_wr),
+
+      .address(addr_core_reg),
+      .write_data(sys_write_data),
+      .read_data(read_data_{core.instance_name}){core.error_port}
+      );
+
+{core.one_cycle_delay}
+
+"""
+
+# Template used by ModExpS6Core.createInstance().  This is different
+# enough from the base template that it's easier to make this separate.
+
+createInstance_template_ModExpS6 = """\
+   //----------------------------------------------------------------
+   // {core.upper_instance_name}
+   //----------------------------------------------------------------
+   wire                 enable_{core.instance_name} = (addr_core_num >= CORE_ADDR_{core.upper_instance_name}) && (addr_core_num <= CORE_ADDR_{core.upper_instance_name} + 9'h03);
+   wire [31: 0]         read_data_{core.instance_name};
+   wire [1:0]           {core.instance_name}_prefix = addr_core_num[1:0] - CORE_ADDR_{core.upper_instance_name};
+
+   {core.name}_wrapper {core.instance_name}_inst
+     (
+      .clk(sys_clk),
+      {core.reset_pin},
+
+      .cs(enable_{core.instance_name} & (sys_eim_rd | sys_eim_wr)),
+      .we(sys_eim_wr),
+
+      .address({{{core.instance_name}_prefix, addr_core_reg}}),
+      .write_data(sys_write_data),
+      .read_data(read_data_{core.instance_name})
+      );
+
+
+"""
+
+# Template used by TRNGCore.createInstance(); this is different enough
+# from the generic template that it's (probably) clearer to have this
+# separate.
+
+createInstance_template_TRNG = """\
+   //----------------------------------------------------------------
+   // {core.upper_instance_name}
+   //----------------------------------------------------------------
+   wire                 enable_{core.instance_name} = (addr_core_num >= CORE_ADDR_{core.upper_instance_name}) && (addr_core_num <= CORE_ADDR_{core.upper_instance_name} + 9'h0f);
+   wire [31: 0]         read_data_{core.instance_name};
+   wire                 error_{core.instance_name};
+   wire [3:0]           {core.instance_name}_prefix = addr_core_num[3:0] - CORE_ADDR_{core.upper_instance_name};
+
+   {core.name} {core.instance_name}_inst
+     (
+      .clk(sys_clk),
+      {core.reset_pin},
+
+      .cs(enable_{core.instance_name} & (sys_eim_rd | sys_eim_wr)),
+      .we(sys_eim_wr),
+
+      .address({{{core.instance_name}_prefix, addr_core_reg}}),
+      .write_data(sys_write_data),
+      .read_data(read_data_{core.instance_name}),
+      .error(error_{core.instance_name}),
+
+      .avalanche_noise(noise),
+      .debug(debug)
+      );
+
+{core.one_cycle_delay}
+
+"""
+
+# Template for one-cycle delay code.
+
+one_cycle_delay_template = """\
+   reg  [31: 0] read_data_{core.instance_name}_reg;
+   always @(posedge sys_clk)
+     read_data_{core.instance_name}_reg <= read_data_{core.instance_name};
+"""
+
+# Template for .createMux() methods.
+
+createMux_template = """\
+       CORE_ADDR_{core.upper_instance_name}:
+         begin
+            sys_read_data_mux = {core0.mux_data_reg};
+            sys_error_mux = {core0.mux_error_reg};
+         end
+"""
+
+# Template for ModExpS6.createMux() method.
+
+createMux_modexps6_template = """\
+       CORE_ADDR_{core.upper_instance_name} + 0,
+       CORE_ADDR_{core.upper_instance_name} + 1,
+       CORE_ADDR_{core.upper_instance_name} + 2,
+       CORE_ADDR_{core.upper_instance_name} + 3:
+         begin
+            sys_read_data_mux = {core0.mux_data_reg};
+            sys_error_mux = {core0.mux_error_reg};
+         end
+"""
+
+
+# Top-level (createModule) template.
+
+createModule_template = """\
+// NOTE: This file is generated; do not edit by hand.
+
+module core_selector
+  (
+   input wire          sys_clk,
+   input wire          sys_rst,
+
+   input wire [16: 0]  sys_eim_addr,
+   input wire          sys_eim_wr,
+   input wire          sys_eim_rd,
+   output wire [31: 0] sys_read_data,
+   input wire [31: 0]  sys_write_data,
+   output wire         sys_error,
+
+   input wire          noise,
+   output wire [7 : 0] debug
+   );
+
+
+   //----------------------------------------------------------------
+   // Address Decoder
+   //----------------------------------------------------------------
+   // upper 9 bits specify core being addressed
+   wire [ 8: 0]         addr_core_num   = sys_eim_addr[16: 8];
+   // lower 8 bits specify register offset in core
+   wire [ 7: 0]         addr_core_reg   = sys_eim_addr[ 7: 0];
+
+
+   //----------------------------------------------------------------
+   // Core Address Table
+   //----------------------------------------------------------------
+{addrs}
+
+{insts}
+   //----------------------------------------------------------------
+   // Output (Read Data) Multiplexer
+   //----------------------------------------------------------------
+   reg [31: 0]          sys_read_data_mux;
+   assign               sys_read_data = sys_read_data_mux;
+   reg                  sys_error_mux;
+   assign               sys_error = sys_error_mux;
+
+   always @*
+
+     case (addr_core_num)
+{muxes}
+       default:
+         begin
+            sys_read_data_mux = {{32{{1'b0}}}};
+            sys_error_mux = 1;
+         end
+     endcase
+
+
+endmodule
+
+
+//======================================================================
+// EOF core_selector.v
+//======================================================================
+"""
+
+# Template for makefile snippet listing Verilog source files.
+
+listVfiles_template = """\
+# NOTE: This file is generated; do not edit by hand.
+
+vfiles +={vfiles}
+"""
+
+# Run main program.
+
+if __name__ == "__main__":
+    main()