diff options
author | Gerhard Sittig <gerhard.sittig@gmx.net> | 2020-06-20 09:31:18 +0200 |
---|---|---|
committer | Gerhard Sittig <gerhard.sittig@gmx.net> | 2020-06-24 19:11:32 +0200 |
commit | 65a1134ade0ca1cd1f2789819a239cee3ca90b21 (patch) | |
tree | 1fcb6b60d1c01e18b2d4ee9a06e94ccb287b12b1 | |
parent | db18ba49bcbe38a9a67297207d821bf0e1e9dc4c (diff) | |
download | libsigrokdecode-65a1134ade0ca1cd1f2789819a239cee3ca90b21.tar.gz libsigrokdecode-65a1134ade0ca1cd1f2789819a239cee3ca90b21.zip |
pjon: introduce decoder for PJON frame inspection
Introduce a protocol decoder which accepts 'pjon-link' Python input and
interprets PJON frames. The implementation is assumed to be operational
but most of the protocol's flexibility (optionally present and variable
width fields) has not yet been tested due to lack of example captures.
During development of the PJON decoder only the PJDL link layer decoder
was available, other link layers were not tested.
-rw-r--r-- | decoders/pjon/__init__.py | 25 | ||||
-rw-r--r-- | decoders/pjon/pd.py | 517 |
2 files changed, 542 insertions, 0 deletions
diff --git a/decoders/pjon/__init__.py b/decoders/pjon/__init__.py new file mode 100644 index 0000000..579fb59 --- /dev/null +++ b/decoders/pjon/__init__.py @@ -0,0 +1,25 @@ +## +## This file is part of the libsigrokdecode project. +## +## Copyright (C) 2020 Gerhard Sittig <gerhard.sittig@gmx.net> +## +## 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, see <http://www.gnu.org/licenses/>. +## + +''' +This protocol decoder interprets the PJON protocol on top of the PJDL +link layer (and potentially other link layers). +''' + +from .pd import Decoder diff --git a/decoders/pjon/pd.py b/decoders/pjon/pd.py new file mode 100644 index 0000000..cf03ef6 --- /dev/null +++ b/decoders/pjon/pd.py @@ -0,0 +1,517 @@ +## +## This file is part of the libsigrokdecode project. +## +## Copyright (C) 2020 Gerhard Sittig <gerhard.sittig@gmx.net> +## +## 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, see <http://www.gnu.org/licenses/>. +## + +# See the https://www.pjon.org/ PJON project page and especially the +# https://www.pjon.org/PJON-protocol-specification-v3.2.php protocol +# specification, which can use different link layers. + +# TODO +# - Handle RX and TX identifiers in separate routines, including the +# optional bus identifers. Track those addresses, add formatters for +# them, and emit "communication relation" details. +# - Check for correct endianess in variable width fields. The spec says +# "network order", that's what this implementation uses. +# - Check for the correct order of optional fields (the spec is not as +# explicit on these details as I'd expect). +# - Check decoder's robustness, completeness, and correctness when more +# captures become available. Currently there are only few, which only +# cover minimal communication, and none of the protocol's flexibility. +# The decoder was essentially written based on the available docs, and +# then took some arbitrary choices and liberties to cope with real life +# data from an example setup. Strictly speaking this decoder violates +# the spec, and errs towards the usability side. + +import sigrokdecode as srd +import struct + +ANN_RX_INFO, ANN_HDR_CFG, ANN_PKT_LEN, ANN_META_CRC, ANN_TX_INFO, \ +ANN_SVC_ID, ANN_PKT_ID, ANN_ANON_DATA, ANN_PAYLOAD, ANN_END_CRC, \ +ANN_SYN_RSP, \ +ANN_RELATION, \ +ANN_WARN, \ + = range(13) + +def calc_crc8(data): + crc = 0 + for b in data: + crc ^= b + for i in range(8): + odd = crc % 2 + crc >>= 1 + if odd: + crc ^= 0x97 + return crc + +def calc_crc32(data): + crc = 0xffffffff + for b in data: + crc ^= b + for i in range(8): + odd = crc % 2 + crc >>= 1 + if odd: + crc ^= 0xedb88320 + crc ^= 0xffffffff + return crc + +class Decoder(srd.Decoder): + api_version = 3 + id = 'pjon' + name = 'PJON' + longname = 'PJON' + desc = 'The PJON protocol.' + license = 'gplv2+' + inputs = ['pjon-link'] + outputs = [] + tags = ['Embedded'] + annotations = ( + ('rx_info', 'Receiver ID'), + ('hdr_cfg', 'Header config'), + ('pkt_len', 'Packet length'), + ('meta_crc', 'Meta CRC'), + ('tx_info', 'Sender ID'), + ('port', 'Service ID'), + ('pkt_id', 'Packet ID'), + ('anon', 'Anonymous data'), + ('payload', 'Payload'), + ('end_crc', 'End CRC'), + ('syn_rsp', 'Sync response'), + ('relation', 'Relation'), + ('warning', 'Warning'), + ) + annotation_rows = ( + ('fields', 'Fields', ( + ANN_RX_INFO, ANN_HDR_CFG, ANN_PKT_LEN, ANN_META_CRC, ANN_TX_INFO, + ANN_SVC_ID, ANN_ANON_DATA, ANN_PAYLOAD, ANN_END_CRC, ANN_SYN_RSP, + )), + ('relations', 'Relations', (ANN_RELATION,)), + ('warnings', 'Warnings', (ANN_WARN,)), + ) + + def __init__(self): + self.reset() + + def reset(self): + self.reset_frame() + + def reset_frame(self): + self.frame_bytes = None + self.ack_bytes = None + self.ann_ss = None + self.ann_es = None + + def start(self): + self.out_ann = self.register(srd.OUTPUT_ANN) + + def putg(self, ss, es, ann, data): + self.put(ss, es, self.out_ann, [ann, data]) + + def frame_flush(self): + if not self.frame_bytes: + return + # TODO Emit "communication relation" details here? + + def handle_field_get_desc(self, idx = None): + '''Lookup description of a PJON frame field.''' + if not self.field_desc: + return None + if idx is None: + idx = self.field_desc_idx + if idx >= 0 and idx >= len(self.field_desc): + return None + if idx < 0 and abs(idx) > len(self.field_desc): + return None + desc = self.field_desc[idx] + return desc + + def handle_field_add_desc(self, fmt, hdl, cls = None): + '''Register description for a PJON frame field.''' + item = { + 'format': fmt, + 'width': struct.calcsize(fmt), + 'handler': hdl, + 'anncls': cls, + } + self.field_desc.append(item) + + def handle_field_seed_desc(self): + '''Seed list of PJON frame fields' descriptions.''' + + # At the start of a PJON frame, the layout of only two fields + # is known. Subsequent fields (their presence, and width) depend + # on the content of the header config field. + + self.field_desc = [] + self.handle_field_add_desc('<B', self.handle_field_rx_id, ANN_RX_INFO) + self.handle_field_add_desc('<B', self.handle_field_config, ANN_HDR_CFG) + + self.field_desc_idx = 0 + self.field_desc_got = 0 + + self.frame_rx_id = None + self.frame_is_broadcast = None + self.frame_tx_id = None + self.frame_payload = None + self.frame_has_ack = None + + def handle_field_rx_id(self, b): + '''Process receiver ID field of a PJON frame.''' + + b = b[0] + + # Track RX info for communication relation emission. + self.frame_rx_id = b + self.frame_is_broadcast = b == 0 + + # Provide text presentation, caller emits frame field annotation. + if b == 255: # "not assigned" + id_txt = 'NA' + elif b == 0: # "broadcast" + id_txt = 'BC' + else: # unicast + id_txt = '{:d}'.format(b) + texts = [ + 'RX_ID {}'.format(id_txt), + '{}'.format(id_txt), + ] + return texts + + def handle_field_config(self, b): + '''Process header config field of a PJON frame.''' + + # Caller provides a list of values. We want a single scalar. + b = b[0] + + # Get the config flags. + self.cfg_shared = b & (1 << 0) + self.cfg_tx_info = b & (1 << 1) + self.cfg_sync_ack = b & (1 << 2) + self.cfg_async_ack = b & (1 << 3) + self.cfg_port = b & (1 << 4) + self.cfg_crc32 = b & (1 << 5) + self.cfg_len16 = b & (1 << 6) + self.cfg_pkt_id = b & (1 << 7) + + # Get a textual presentation of the flags. + text = [] + text.append('pkt_id' if self.cfg_pkt_id else '-') # packet number + text.append('len16' if self.cfg_len16 else '-') # 16bit length not 8bit + text.append('crc32' if self.cfg_crc32 else '-') # 32bit CRC not 8bit + text.append('svc_id' if self.cfg_port else '-') # port aka service ID + text.append('ack_mode' if self.cfg_async_ack else '-') # async response + text.append('ack' if self.cfg_sync_ack else '-') # synchronous response + text.append('tx_info' if self.cfg_tx_info else '-') + text.append('bus_id' if self.cfg_shared else '-') # "shared" vs "local" + text = ' '.join(text) + bits = '{:08b}'.format(b) + texts = [ + 'CFG {:s}'.format(text), + 'CFG {}'.format(bits), + bits + ] + + # TODO Come up with the most appropriate phrases for this logic. + # Are separate instruction groups with repeated conditions more + # readable than one common block which registers fields _and_ + # updates the overhead size? Or is the latter preferrable due to + # easier maintenance and less potential for inconsistency? + + # Get the size of variable width fields, to calculate the size + # of the packet overhead (the part that is not the payload data). + # This lets us derive the payload length when we later receive + # the packet length. + u8_fmt = '>B' + u16_fmt = '>H' + u32_fmt = '>L' + len_fmt = u16_fmt if self.cfg_len16 else u8_fmt + crc_fmt = u32_fmt if self.cfg_crc32 else u8_fmt + self.cfg_overhead = 0 + self.cfg_overhead += struct.calcsize(u8_fmt) # receiver ID + self.cfg_overhead += struct.calcsize(u8_fmt) # header config + self.cfg_overhead += struct.calcsize(len_fmt) # packet length + self.cfg_overhead += struct.calcsize(u8_fmt) # initial CRC, always CRC8 + # TODO Check for completeness and correctness. + if self.cfg_shared: + self.cfg_overhead += struct.calcsize(u32_fmt) # receiver bus + if self.cfg_tx_info: + if self.cfg_shared: + self.cfg_overhead += struct.calcsize(u32_fmt) # sender bus + self.cfg_overhead += struct.calcsize(u8_fmt) # sender ID + if self.cfg_port: + self.cfg_overhead += struct.calcsize(u16_fmt) # service ID + if self.cfg_pkt_id: + self.cfg_overhead += struct.calcsize(u16_fmt) # packet ID + self.cfg_overhead += struct.calcsize(crc_fmt) # end CRC + + # Register more frame fields as we learn about their presence and + # format. Up to this point only receiver ID and header config were + # registered since their layout is fixed. + # + # Packet length and meta CRC are always present but can be of + # variable width. Optional fields follow the meta CRC and preceed + # the payload bytes. Notice that payload length isn't known here + # either, though its position is known already. The packet length + # is yet to get received. Subtracting the packet overhead from it + # (which depends on the header configuration) will provide that + # information. + # + # TODO Check for completeness and correctness. + # TODO Optionally fold overhead size arith and field registration + # into one block of instructions, to reduce the redundancy in the + # condition checks, and raise awareness for incomplete sequences + # during maintenance. + self.handle_field_add_desc(len_fmt, self.handle_field_pkt_len, ANN_PKT_LEN) + self.handle_field_add_desc(u8_fmt, self.handle_field_init_crc, ANN_META_CRC) + if self.cfg_shared: + self.handle_field_add_desc(u32_fmt, ['RX_BUS {:08x}', '{:08x}'], ANN_ANON_DATA) + if self.cfg_tx_info: + if self.cfg_shared: + self.handle_field_add_desc(u32_fmt, ['TX_BUS {:08x}', '{:08x}'], ANN_ANON_DATA) + self.handle_field_add_desc(u8_fmt, ['TX_ID {:d}', '{:d}'], ANN_ANON_DATA) + if self.cfg_port: + self.handle_field_add_desc(u16_fmt, ['PORT {:d}', '{:d}'], ANN_ANON_DATA) + if self.cfg_pkt_id: + self.handle_field_add_desc(u16_fmt, ['PKT {:04x}', '{:04x}'], ANN_ANON_DATA) + pl_fmt = '>{:d}B'.format(0) + self.handle_field_add_desc(pl_fmt, self.handle_field_payload, ANN_PAYLOAD) + self.handle_field_add_desc(crc_fmt, self.handle_field_end_crc, ANN_END_CRC) + + # Emit warning annotations for invalid flag combinations. + warn_texts = [] + wants_ack = self.cfg_sync_ack or self.cfg_async_ack + if wants_ack and not self.cfg_tx_info: + warn_texts.append('ACK request without TX info') + if wants_ack and self.frame_is_broadcast: + warn_texts.append('ACK request for broadcast') + if self.cfg_sync_ack and self.cfg_async_ack: + warn_texts.append('sync and async ACK request') + if self.cfg_len16 and not self.cfg_crc32: + warn_texts.append('extended length needs CRC32') + if warn_texts: + warn_texts = ', '.join(warn_texts) + self.putg(self.ann_ss, self.ann_es, ANN_WARN, [warn_texts]) + + # Have the caller emit the annotation for configuration data. + return texts + + def handle_field_pkt_len(self, b): + '''Process packet length field of a PJON frame.''' + + # Caller provides a list of values. We want a single scalar. + b = b[0] + + # The wire communicates the total packet length. Some of it is + # overhead (non-payload data), while its volume is variable in + # size (dpends on the header configuration). + # + # Derive the payload size from previously observed flags. Update + # the previously registered field description (the second last + # item in the list, before the end CRC). + + pkt_len = b + pl_len = b - self.cfg_overhead + warn_texts = [] + if pkt_len not in range(self.cfg_overhead, 65536): + warn_texts.append('suspicious packet length') + if pkt_len > 15 and not self.cfg_crc32: + warn_texts.append('length above 15 needs CRC32') + if pl_len < 1: + warn_texts.append('suspicious payload length') + pl_len = 0 + if warn_texts: + warn_texts = ', '.join(warn_texts) + self.putg(self.ann_ss, self.ann_es, ANN_WARN, [warn_texts]) + pl_fmt = '>{:d}B'.format(pl_len) + + desc = self.handle_field_get_desc(-2) + desc['format'] = pl_fmt + desc['width'] = struct.calcsize(pl_fmt) + + # Have the caller emit the annotation for the packet length. + # Provide information of different detail level for zooming. + texts = [ + 'LENGTH {:d} (PAYLOAD {:d})'.format(pkt_len, pl_len), + 'LEN {:d} (PL {:d})'.format(pkt_len, pl_len), + '{:d} ({:d})'.format(pkt_len, pl_len), + '{:d}'.format(pkt_len), + ] + return texts + + def handle_field_common_crc(self, have, is_meta): + '''Process a CRC field of a PJON frame.''' + + # CRC algorithm and width are configurable, and can differ + # across meta and end checksums in a frame's fields. + caption = 'META' if is_meta else 'END' + crc_len = 8 if is_meta else 32 if self.cfg_crc32 else 8 + crc_fmt = '{:08x}' if crc_len == 32 else '{:02x}' + have_text = crc_fmt.format(have) + + # Check received against expected checksum. Emit warnings. + warn_texts = [] + data = self.frame_bytes[:-1] + want = calc_crc32(data) if crc_len == 32 else calc_crc8(data) + if want != have: + want_text = crc_fmt.format(want) + warn_texts.append('CRC mismatch - want {} have {}'.format(want_text, have_text)) + if warn_texts: + warn_texts = ', '.join(warn_texts) + self.putg(self.ann_ss, self.ann_es, ANN_WARN, [warn_texts]) + + # Provide text representation for frame field, caller emits + # the annotation. + texts = [ + '{}_CRC {}'.format(caption, have_text), + 'CRC {}'.format(have_text), + have_text, + ] + return texts + + def handle_field_init_crc(self, b): + '''Process initial CRC (meta) field of a PJON frame.''' + # Caller provides a list of values. We want a single scalar. + b = b[0] + return self.handle_field_common_crc(b, True) + + def handle_field_end_crc(self, b): + '''Process end CRC (total frame) field of a PJON frame.''' + # Caller provides a list of values. We want a single scalar. + b = b[0] + return self.handle_field_common_crc(b, False) + + def handle_field_payload(self, b): + '''Process payload data field of a PJON frame.''' + + self.frame_payload = b[:] + + text = ' '.join(['{:02x}'.format(v) for v in b]) + texts = [ + 'PAYLOAD {}'.format(text), + text, + ] + return texts + + def handle_field_sync_resp(self, b): + '''Process synchronous response for a PJON frame.''' + + self.frame_has_ack = b + + texts = [ + 'ACK {:02x}'.format(b), + '{:02x}'.format(b), + ] + return texts + + def decode(self, ss, es, data): + ptype, pdata = data + + # Start frame bytes accumulation when FRAME_INIT is seen. Flush + # previously accumulated frame bytes when a new frame starts. + if ptype == 'FRAME_INIT': + self.frame_flush() + self.reset_frame() + self.frame_bytes = [] + self.handle_field_seed_desc() + return + + # Use IDLE as another (earlier) trigger to flush frames. Also + # trigger flushes on FRAME-DATA which mean that the link layer + # inspection has seen the end of a protocol frame. + # + # TODO Improve usability? Emit warnings for PJON frames where + # FRAME_DATA was seen but FRAME_INIT wasn't? So that users can + # become aware of broken frames. + if ptype in ('IDLE', 'FRAME_DATA'): + self.frame_flush() + self.reset_frame() + return + + # Switch from data bytes to response bytes when WAIT is seen. + if ptype == 'SYNC_RESP_WAIT': + self.ack_bytes = [] + self.ann_ss, self.ann_es = None, None + return + + # Accumulate data bytes as they arrive. Put them in the bucket + # which corresponds to its most recently seen leader. + if ptype == 'DATA_BYTE': + b = pdata + + # Are we collecting response bytes (ACK)? + if self.ack_bytes is not None: + if not self.ann_ss: + self.ann_ss = ss + self.ack_bytes.append(b) + self.ann_es = es + text = self.handle_field_sync_resp(b) + if text: + self.putg(self.ann_ss, self.ann_es, ANN_SYN_RSP, text) + self.ann_ss, self.ann_es = None, None + return + + # Are we collecting frame content? + if self.frame_bytes is not None: + if not self.ann_ss: + self.ann_ss = ss + self.frame_bytes.append(b) + self.ann_es = es + + # Has the field value become available yet? + desc = self.handle_field_get_desc() + if not desc: + return + width = desc.get('width', None) + if not width: + return + self.field_desc_got += 1 + if self.field_desc_got != width: + return + + # Grab most recent received field as a byte array. Get + # the values that it contains. + fmt = desc.get('format', '>B') + raw = bytearray(self.frame_bytes[-width:]) + values = struct.unpack(fmt, raw) + + # Process the value, and get its presentation. Can be + # mere formatting, or serious execution of logic. + hdl = desc.get('handler', '{!r}') + if isinstance(hdl, str): + text = [hdl.format(*values)] + elif isinstance(hdl, (list, tuple)): + text = [f.format(*values) for f in hdl] + elif hdl: + text = hdl(values) + cls = desc.get('anncls', ANN_ANON_DATA) + + # Emit annotation unless the handler routine already did. + if cls is not None and text: + self.putg(self.ann_ss, self.ann_es, cls, text) + self.ann_ss, self.ann_es = None, None + + # Advance scan position for to-get-received field. + self.field_desc_idx += 1 + self.field_desc_got = 0 + return + + # Unknown phase, not collecting. Not synced yet to the input? + return + + # Unknown or unhandled kind of link layer output. + return |