## ## This file is part of the libsigrokdecode project. ## ## Copyright (C) 2011-2014 Uwe Hermann ## ## 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 ## the Free Software Foundation; either version 2 of the License, or ## (at your option) any later version. ## ## This program is distributed in the hope that it will be useful, ## but WITHOUT ANY WARRANTY; without even the implied warranty of ## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the ## GNU General Public License for more details. ## ## You should have received a copy of the GNU General Public License ## along with this program; if not, write to the Free Software ## Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA ## import sigrokdecode as srd ''' OUTPUT_PYTHON format: Packet: [, , ] This is the list of s and their respective values: - 'STARTBIT': The data is the (integer) value of the start bit (0/1). - ('DATA', 'DATABITS'): DATA is the (integer) value of the UART data. Valid values range from 0 to 512 (as the data can be up to 9 bits in size). 'DATABITS' is the list of individual data bits and their ss/es numbers. - 'PARITYBIT': The data is the (integer) value of the parity bit (0/1). - 'STOPBIT': The data is the (integer) value of the stop bit (0 or 1). - 'INVALID STARTBIT': The data is the (integer) value of the start bit (0/1). - 'INVALID STOPBIT': The data is the (integer) value of the stop bit (0/1). - 'PARITY ERROR': The data is a tuple with two entries. The first one is the expected parity value, the second is the actual parity value. - TODO: Frame error? The field is 0 for RX packets, 1 for TX packets. ''' # Used for differentiating between the two data directions. RX = 0 TX = 1 # Given a parity type to check (odd, even, zero, one), the value of the # parity bit, the value of the data, and the length of the data (5-9 bits, # usually 8 bits) return True if the parity is correct, False otherwise. # 'none' is _not_ allowed as value for 'parity_type'. def parity_ok(parity_type, parity_bit, data, num_data_bits): # Handle easy cases first (parity bit is always 1 or 0). if parity_type == 'zero': return parity_bit == 0 elif parity_type == 'one': return parity_bit == 1 # Count number of 1 (high) bits in the data (and the parity bit itself!). ones = bin(data).count('1') + parity_bit # Check for odd/even parity. if parity_type == 'odd': return (ones % 2) == 1 elif parity_type == 'even': return (ones % 2) == 0 class SamplerateError(Exception): pass class ChannelError(Exception): pass class Decoder(srd.Decoder): api_version = 2 id = 'uart' name = 'UART' longname = 'Universal Asynchronous Receiver/Transmitter' desc = 'Asynchronous, serial bus.' license = 'gplv2+' inputs = ['logic'] outputs = ['uart'] optional_channels = ( # Allow specifying only one of the signals, e.g. if only one data # direction exists (or is relevant). {'id': 'rx', 'name': 'RX', 'desc': 'UART receive line'}, {'id': 'tx', 'name': 'TX', 'desc': 'UART transmit line'}, ) options = ( {'id': 'baudrate', 'desc': 'Baud rate', 'default': 115200}, {'id': 'num_data_bits', 'desc': 'Data bits', 'default': 8, 'values': (5, 6, 7, 8, 9)}, {'id': 'parity_type', 'desc': 'Parity type', 'default': 'none', 'values': ('none', 'odd', 'even', 'zero', 'one')}, {'id': 'parity_check', 'desc': 'Check parity?', 'default': 'yes', 'values': ('yes', 'no')}, {'id': 'num_stop_bits', 'desc': 'Stop bits', 'default': 1.0, 'values': (0.0, 0.5, 1.0, 1.5)}, {'id': 'bit_order', 'desc': 'Bit order', 'default': 'lsb-first', 'values': ('lsb-first', 'msb-first')}, {'id': 'format', 'desc': 'Data format', 'default': 'ascii', 'values': ('ascii', 'dec', 'hex', 'oct', 'bin')}, {'id': 'invert_rx', 'desc': 'Invert RX?', 'default': 'no', 'values': ('yes', 'no')}, {'id': 'invert_tx', 'desc': 'Invert TX?', 'default': 'no', 'values': ('yes', 'no')}, ) annotations = ( ('rx-data', 'RX data'), ('tx-data', 'TX data'), ('rx-start', 'RX start bits'), ('tx-start', 'TX start bits'), ('rx-parity-ok', 'RX parity OK bits'), ('tx-parity-ok', 'TX parity OK bits'), ('rx-parity-err', 'RX parity error bits'), ('tx-parity-err', 'TX parity error bits'), ('rx-stop', 'RX stop bits'), ('tx-stop', 'TX stop bits'), ('rx-warnings', 'RX warnings'), ('tx-warnings', 'TX warnings'), ('rx-data-bits', 'RX data bits'), ('tx-data-bits', 'TX data bits'), ) annotation_rows = ( ('rx-data', 'RX', (0, 2, 4, 6, 8)), ('rx-data-bits', 'RX bits', (12,)), ('rx-warnings', 'RX warnings', (10,)), ('tx-data', 'TX', (1, 3, 5, 7, 9)), ('tx-data-bits', 'TX bits', (13,)), ('tx-warnings', 'TX warnings', (11,)), ) binary = ( ('rx', 'RX dump'), ('tx', 'TX dump'), ('rxtx', 'RX/TX dump'), ) def putx(self, rxtx, data): s, halfbit = self.startsample[rxtx], int(self.bit_width / 2) self.put(s - halfbit, self.samplenum + halfbit, self.out_ann, data) def putpx(self, rxtx, data): s, halfbit = self.startsample[rxtx], int(self.bit_width / 2) self.put(s - halfbit, self.samplenum + halfbit, self.out_python, data) def putg(self, data): s, halfbit = self.samplenum, int(self.bit_width / 2) self.put(s - halfbit, s + halfbit, self.out_ann, data) def putp(self, data): s, halfbit = self.samplenum, int(self.bit_width / 2) self.put(s - halfbit, s + halfbit, self.out_python, data) def putbin(self, rxtx, data): s, halfbit = self.startsample[rxtx], int(self.bit_width / 2) self.put(s - halfbit, self.samplenum + halfbit, self.out_bin, data) def __init__(self, **kwargs): self.samplerate = None self.samplenum = 0 self.frame_start = [-1, -1] self.startbit = [-1, -1] self.cur_data_bit = [0, 0] self.databyte = [0, 0] self.paritybit = [-1, -1] self.stopbit1 = [-1, -1] self.startsample = [-1, -1] self.state = ['WAIT FOR START BIT', 'WAIT FOR START BIT'] self.oldbit = [1, 1] self.oldpins = [1, 1] self.databits = [[], []] def start(self): self.out_python = self.register(srd.OUTPUT_PYTHON) self.out_bin = self.register(srd.OUTPUT_BINARY) self.out_ann = self.register(srd.OUTPUT_ANN) def metadata(self, key, value): if key == srd.SRD_CONF_SAMPLERATE: self.samplerate = value # The width of one UART bit in number of samples. self.bit_width = float(self.samplerate) / float(self.options['baudrate']) # Return true if we reached the middle of the desired bit, false otherwise. def reached_bit(self, rxtx, bitnum): # bitpos is the samplenumber which is in the middle of the # specified UART bit (0 = start bit, 1..x = data, x+1 = parity bit # (if used) or the first stop bit, and so on). bitpos = self.frame_start[rxtx] + (self.bit_width / 2.0) bitpos += bitnum * self.bit_width if self.samplenum >= bitpos: return True return False def reached_bit_last(self, rxtx, bitnum): bitpos = self.frame_start[rxtx] + ((bitnum + 1) * self.bit_width) if self.samplenum >= bitpos: return True return False def wait_for_start_bit(self, rxtx, old_signal, signal): # The start bit is always 0 (low). As the idle UART (and the stop bit) # level is 1 (high), the beginning of a start bit is a falling edge. if not (old_signal == 1 and signal == 0): return # Save the sample number where the start bit begins. self.frame_start[rxtx] = self.samplenum self.state[rxtx] = 'GET START BIT' def get_start_bit(self, rxtx, signal): # Skip samples until we're in the middle of the start bit. if not self.reached_bit(rxtx, 0): return self.startbit[rxtx] = signal # The startbit must be 0. If not, we report an error. if self.startbit[rxtx] != 0: self.putp(['INVALID STARTBIT', rxtx, self.startbit[rxtx]]) # TODO: Abort? Ignore rest of the frame? self.cur_data_bit[rxtx] = 0 self.databyte[rxtx] = 0 self.startsample[rxtx] = -1 self.state[rxtx] = 'GET DATA BITS' self.putp(['STARTBIT', rxtx, self.startbit[rxtx]]) self.putg([rxtx + 2, ['Start bit', 'Start', 'S']]) def get_data_bits(self, rxtx, signal): # Skip samples until we're in the middle of the desired data bit. if not self.reached_bit(rxtx, self.cur_data_bit[rxtx] + 1): return # Save the sample number of the middle of the first data bit. if self.startsample[rxtx] == -1: self.startsample[rxtx] = self.samplenum # Get the next data bit in LSB-first or MSB-first fashion. if self.options['bit_order'] == 'lsb-first': self.databyte[rxtx] >>= 1 self.databyte[rxtx] |= \ (signal << (self.options['num_data_bits'] - 1)) else: self.databyte[rxtx] <<= 1 self.databyte[rxtx] |= (signal << 0) self.putg([rxtx + 12, ['%d' % signal]]) # Store individual data bits and their start/end samplenumbers. s, halfbit = self.samplenum, int(self.bit_width / 2) self.databits[rxtx].append([signal, s - halfbit, s + halfbit]) # Return here, unless we already received all data bits. if self.cur_data_bit[rxtx] < self.options['num_data_bits'] - 1: self.cur_data_bit[rxtx] += 1 return self.state[rxtx] = 'GET PARITY BIT' self.putpx(rxtx, ['DATA', rxtx, (self.databyte[rxtx], self.databits[rxtx])]) b, f = self.databyte[rxtx], self.options['format'] if f == 'ascii': c = chr(b) if b in range(30, 126 + 1) else '[%02X]' % b self.putx(rxtx, [rxtx, [c]]) elif f == 'dec': self.putx(rxtx, [rxtx, [str(b)]]) elif f == 'hex': self.putx(rxtx, [rxtx, [hex(b)[2:].zfill(2).upper()]]) elif f == 'oct': self.putx(rxtx, [rxtx, [oct(b)[2:].zfill(3)]]) elif f == 'bin': self.putx(rxtx, [rxtx, [bin(b)[2:].zfill(8)]]) self.putbin(rxtx, (rxtx, bytes([b]))) self.putbin(rxtx, (2, bytes([b]))) self.databits = [[], []] def get_parity_bit(self, rxtx, signal): # If no parity is used/configured, skip to the next state immediately. if self.options['parity_type'] == 'none': self.state[rxtx] = 'GET STOP BITS' return # Skip samples until we're in the middle of the parity bit. if not self.reached_bit(rxtx, self.options['num_data_bits'] + 1): return self.paritybit[rxtx] = signal self.state[rxtx] = 'GET STOP BITS' if parity_ok(self.options['parity_type'], self.paritybit[rxtx], self.databyte[rxtx], self.options['num_data_bits']): self.putp(['PARITYBIT', rxtx, self.paritybit[rxtx]]) self.putg([rxtx + 4, ['Parity bit', 'Parity', 'P']]) else: # TODO: Return expected/actual parity values. self.putp(['PARITY ERROR', rxtx, (0, 1)]) # FIXME: Dummy tuple... self.putg([rxtx + 6, ['Parity error', 'Parity err', 'PE']]) # TODO: Currently only supports 1 stop bit. def get_stop_bits(self, rxtx, signal): # Skip samples until we're in the middle of the stop bit(s). skip_parity = 0 if self.options['parity_type'] == 'none' else 1 b = self.options['num_data_bits'] + 1 + skip_parity if not self.reached_bit(rxtx, b): return self.stopbit1[rxtx] = signal # Stop bits must be 1. If not, we report an error. if self.stopbit1[rxtx] != 1: self.putp(['INVALID STOPBIT', rxtx, self.stopbit1[rxtx]]) self.putg([rxtx + 8, ['Frame error', 'Frame err', 'FE']]) # TODO: Abort? Ignore the frame? Other? self.state[rxtx] = 'WAIT FOR START BIT' self.putp(['STOPBIT', rxtx, self.stopbit1[rxtx]]) self.putg([rxtx + 4, ['Stop bit', 'Stop', 'T']]) def decode(self, ss, es, data): if not self.samplerate: raise SamplerateError('Cannot decode without samplerate.') for (self.samplenum, pins) in data: # Note: Ignoring identical samples here for performance reasons # is not possible for this PD, at least not in the current state. # if self.oldpins == pins: # continue self.oldpins, (rx, tx) = pins, pins if self.options['invert_rx'] == 'yes': rx = not rx if self.options['invert_tx'] == 'yes': tx = not tx # Either RX or TX (but not both) can be omitted. has_pin = [rx in (0, 1), tx in (0, 1)] if has_pin == [False, False]: raise ChannelError('Either TX or RX (or both) pins required.') # State machine. for rxtx in (RX, TX): # Don't try to handle RX (or TX) if not supplied. if not has_pin[rxtx]: continue signal = rx if (rxtx == RX) else tx if self.state[rxtx] == 'WAIT FOR START BIT': self.wait_for_start_bit(rxtx, self.oldbit[rxtx], signal) elif self.state[rxtx] == 'GET START BIT': self.get_start_bit(rxtx, signal) elif self.state[rxtx] == 'GET DATA BITS': self.get_data_bits(rxtx, signal) elif self.state[rxtx] == 'GET PARITY BIT': self.get_parity_bit(rxtx, signal) elif self.state[rxtx] == 'GET STOP BITS': self.get_stop_bits(rxtx, signal) # Save current RX/TX values for the next round. self.oldbit[rxtx] = signal