greaseweazle-firmware/scripts/greaseweazle/image/edsk.py

# greaseweazle/image/edsk.py
#
# Some of the code here is heavily inspired by Simon Owen's SAMdisk:
# https://simonowen.com/samdisk/
#
# Written & released by Keir Fraser <keir.xen@gmail.com>
# 
# This is free and unencumbered software released into the public domain.
# See the file COPYING for more details, or visit <http://unlicense.org>.

import binascii, math, struct
import itertools as it
from bitarray import bitarray

from greaseweazle import error
from greaseweazle.codec.ibm import mfm
from greaseweazle.track import MasterTrack, RawTrack
from .image import Image

class SR1:
    SUCCESS                   = 0x00
    CANNOT_FIND_ID_ADDRESS    = 0x01
    WRITE_PROTECT_DETECTED    = 0x02
    CANNOT_FIND_SECTOR_ID     = 0x04
    RESERVED1                 = 0x08
    OVERRUN                   = 0x10
    CRC_ERROR                 = 0x20
    RESERVED2                 = 0x40
    END_OF_CYLINDER           = 0x80

class SR2:
    SUCCESS                   = 0x00
    MISSING_ADDRESS_MARK      = 0x01
    BAD_CYLINDER              = 0x02
    SCAN_COMMAND_FAILED       = 0x04
    SCAN_COMMAND_EQUAL        = 0x08
    WRONG_CYLINDER_DETECTED   = 0x10
    CRC_ERROR_IN_SECTOR_DATA  = 0x20
    SECTOR_WITH_DELETED_DATA  = 0x40
    RESERVED                  = 0x80

class SectorErrors:
    def __init__(self, sr1, sr2):
        self.id_crc_error = (sr1 & SR1.CRC_ERROR) != 0
        self.data_not_found = (sr2 & SR2.MISSING_ADDRESS_MARK) != 0
        self.data_crc_error = (sr2 & SR2.CRC_ERROR_IN_SECTOR_DATA) != 0
        self.deleted_dam = (sr2 & SR2.SECTOR_WITH_DELETED_DATA) != 0
        if self.data_crc_error:
            # uPD765 sets both id and data flags for data CRC errors
            self.id_crc_error = False
        if (# normal data
            (sr1 == SR1.SUCCESS and sr2 == SR2.SUCCESS) or
            # deleted data
            (sr1 == SR1.SUCCESS and sr2 == SR2.SECTOR_WITH_DELETED_DATA) or
            # end of track
            (sr1 == SR1.END_OF_CYLINDER and sr2 == SR2.SUCCESS) or
            # id crc error
            (sr1 == SR1.CRC_ERROR and sr2 == SR2.SUCCESS) or
            # normal data crc error
            (sr1 == SR1.CRC_ERROR and sr2 == SR2.CRC_ERROR_IN_SECTOR_DATA) or
            # deleted data crc error
            (sr1 == SR1.CRC_ERROR and sr2 == (SR2.CRC_ERROR_IN_SECTOR_DATA |
                                              SR2.SECTOR_WITH_DELETED_DATA)) or
            # data field missing (some FDCs set AM in ST1)
            (sr1 == SR1.CANNOT_FIND_ID_ADDRESS
             and sr2 == SR2.MISSING_ADDRESS_MARK) or
            # data field missing (some FDCs don't)
            (sr1 == SR1.SUCCESS and sr2 == SR2.MISSING_ADDRESS_MARK) or
            # CHRN mismatch
            (sr1 == SR1.CANNOT_FIND_SECTOR_ID and sr2 == SR2.SUCCESS) or
            # CHRN mismatch, including wrong cylinder
            (sr1 == SR1.CANNOT_FIND_SECTOR_ID
             and sr2 == SR2.WRONG_CYLINDER_DETECTED)):
            pass
        else:
            print('Unusual status flags (ST1=%02X ST2=%02X)' % (sr1, sr2))
            
class EDSKTrack:

    gap_presync = 12
    gap_4a = 80 # Post-Index
    gap_1  = 50 # Post-IAM
    gap_2  = 22 # Post-IDAM

    gapbyte = 0x4e
    
    def __init__(self):
        self.time_per_rev = 0.2
        self.clock = 2e-6
        self.bits, self.weak = [], []

    def raw_track(self):
        track = MasterTrack(
            bits = self.bits,
            time_per_rev = self.time_per_rev,
            weak = self.weak)
        track.verify = self
        track.verify_revs = 1
        return track

    def _find_sync(self, bits, sync, start):
        for offs in bits.itersearch(sync):
            if offs >= start:
                return offs
        return None
    
    def verify_track(self, flux):
        flux.cue_at_index()
        raw = RawTrack(clock = self.clock, data = flux)
        bits, _ = raw.get_all_data()
        weak_iter = it.chain(self.weak, [(self.verify_len+1,1)])
        weak = next(weak_iter)

        # Start checking from the IAM sync
        dump_start = self._find_sync(bits, mfm.iam_sync, 0)
        self_start = self._find_sync(self.bits, mfm.iam_sync, 0)

        # Include the IAM pre-sync header
        if dump_start is None:
            return False
        dump_start -= self.gap_presync * 16
        self_start -= self.gap_presync * 16

        while self_start is not None and dump_start is not None:

            # Find the weak areas immediately before and after the current
            # region to be checked.
            s,n = None,None
            while self_start > weak[0]:
                s,n = weak
                weak = next(weak_iter)

            # If there is a weak area preceding us, move the start point to
            # immediately follow the weak area.
            if s is not None:
                delta = self_start - (s + n + 16)
                self_start -= delta
                dump_start -= delta

            # Truncate the region at the next weak area, or the last sector.
            self_end = max(self_start, min(weak[0], self.verify_len+1))
            dump_end = dump_start + self_end - self_start

            # Extract the corresponding areas from the pristine track and
            # from the dump, and check that they match.
            if bits[dump_start:dump_end] != self.bits[self_start:self_end]:
                return False

            # Find the next A1A1A1 sync pattern
            dump_start = self._find_sync(bits, mfm.sync, dump_end)
            self_start = self._find_sync(self.bits, mfm.sync, self_end)

        # Did we verify all regions in the pristine track?
        return self_start is None

class EDSK(Image):

    read_only = True
    default_format = 'ibm.mfm'

    def __init__(self):
        self.to_track = dict()

    # Currently only finds one weak range.
    @staticmethod
    def find_weak_ranges(dat, size):
        orig = dat[:size]
        s, e = size, 0
        for i in range(1, len(dat)//size):
            diff = [x^y for x, y in zip(orig, dat[size*i:size*(i+1)])]
            weak = [idx for idx, val in enumerate(diff) if val != 0]
            if weak:
                s, e = min(s, weak[0]), max(e, weak[-1])
        return [(s,e-s+1)] if s <= e else []

    @staticmethod
    def _build_8k_track(sectors):
        if len(sectors) != 1:
            return None
        c,h,r,n,errs,data = sectors[0]
        if n != 6:
            return None
        if errs.id_crc_error or errs.data_not_found or not errs.data_crc_error:
            return None
        # Magic longtrack value is for Coin-Op Hits. Taken from SAMdisk.
        if len(data) > 6307:
            data = data[:6307]
        track = EDSKTrack()
        t = bytes()
        # Post-index gap
        t += mfm.encode(bytes([track.gapbyte] * 16))
        # IAM
        t += mfm.encode(bytes(track.gap_presync))
        t += mfm.iam_sync_bytes
        t += mfm.encode(bytes([mfm.IBM_MFM.IAM]))
        t += mfm.encode(bytes([track.gapbyte] * 16))
        # IDAM
        t += mfm.encode(bytes(track.gap_presync))
        t += mfm.sync_bytes
        am = bytes([0xa1, 0xa1, 0xa1, mfm.IBM_MFM.IDAM, c, h, r, n])
        crc = mfm.crc16.new(am).crcValue
        am += struct.pack('>H', crc)
        t += mfm.encode(am[3:])
        t += mfm.encode(bytes([track.gapbyte] * track.gap_2))
        # DAM
        t += mfm.encode(bytes(track.gap_presync))
        t += mfm.sync_bytes
        dmark = (mfm.IBM_MFM.DDAM if errs.deleted_dam
                 else mfm.IBM_MFM.DAM)
        am = bytes([0xa1, 0xa1, 0xa1, dmark]) + data
        t += mfm.encode(am[3:])
        # Pre-index gap
        track.verify_len = len(t)*8
        tracklen = int((track.time_per_rev / track.clock) / 16)
        gap = max(40, tracklen - len(t)//2)
        t += mfm.encode(bytes([track.gapbyte] * gap))
        track.bits = bitarray(endian='big')
        track.bits.frombytes(mfm.mfm_encode(t))
        return track

    @classmethod
    def from_file(cls, name):

        with open(name, "rb") as f:
            dat = f.read()

        edsk = cls()

        sig, creator, ncyls, nsides, track_sz = struct.unpack(
            '<34s14s2BH', dat[:52])
        if sig[:8] == b'MV - CPC':
            extended = False
        elif sig[:16] == b'EXTENDED CPC DSK':
            extended = True
        else:
            raise error.Fatal('Unrecognised CPC DSK file: bad signature')

        if extended:
            track_sizes = list(dat[52:52+ncyls*nsides])
            track_sizes = list(map(lambda x: x*256, track_sizes))
        else:
            track_sizes = [track_sz] * (ncyls * nsides)

        o = 256 # skip disk header and track-size table
        for track_size in track_sizes:
            if track_size == 0:
                continue
            sig, cyl, head, sec_sz, nsecs, gap_3, filler = struct.unpack(
                '<12s4x2B2x4B', dat[o:o+24])
            error.check(sig == b'Track-Info\r\n',
                        'EDSK: Missing track header')
            error.check((cyl, head) not in edsk.to_track,
                        'EDSK: Track specified twice')
            bad_crc_clip_data = False
            while True:
                track = EDSKTrack()
                t = bytes()
                # Post-index gap
                t += mfm.encode(bytes([track.gapbyte] * track.gap_4a))
                # IAM
                t += mfm.encode(bytes(track.gap_presync))
                t += mfm.iam_sync_bytes
                t += mfm.encode(bytes([mfm.IBM_MFM.IAM]))
                t += mfm.encode(bytes([track.gapbyte] * track.gap_1))
                sh = dat[o+24:o+24+8*nsecs]
                data_pos = o + 256 # skip track header and sector-info table
                clippable, ngap3, sectors = 0, 0, []
                while sh:
                    c, h, r, n, stat1, stat2, data_size = struct.unpack(
                        '<6BH', sh[:8])
                    sh = sh[8:]
                    native_size = mfm.sec_sz(n)
                    weak = []
                    errs = SectorErrors(stat1, stat2)
                    num_copies = 0 if errs.data_not_found else 1
                    if not extended:
                        data_size = mfm.sec_sz(sec_sz)
                    sec_data = dat[data_pos:data_pos+data_size]
                    data_pos += data_size
                    if (extended
                        and data_size > native_size
                        and errs.data_crc_error
                        and (data_size % native_size == 0
                             or data_size == 49152)):
                        num_copies = (3 if data_size == 49152
                                      else data_size // native_size)
                        data_size //= num_copies
                        weak = cls().find_weak_ranges(sec_data, data_size)
                        sec_data = sec_data[:data_size]
                    sectors.append((c,h,r,n,errs,sec_data))
                    # IDAM
                    t += mfm.encode(bytes(track.gap_presync))
                    t += mfm.sync_bytes
                    am = bytes([0xa1, 0xa1, 0xa1, mfm.IBM_MFM.IDAM,
                                c, h, r, n])
                    crc = mfm.crc16.new(am).crcValue
                    if errs.id_crc_error:
                        crc ^= 0x5555
                    am += struct.pack('>H', crc)
                    t += mfm.encode(am[3:])
                    t += mfm.encode(bytes([track.gapbyte] * track.gap_2))
                    # DAM
                    if errs.id_crc_error or errs.data_not_found:
                        continue
                    t += mfm.encode(bytes(track.gap_presync))
                    t += mfm.sync_bytes
                    track.weak += [((s+len(t)//2+4)*16, n*16) for s,n in weak]
                    dmark = (mfm.IBM_MFM.DDAM if errs.deleted_dam
                             else mfm.IBM_MFM.DAM)
                    gap_included = False
                    if errs.data_crc_error:
                        if sh:
                            # Look for next IDAM
                            idam = bytes([0]*12 + [0xa1]*3
                                         + [mfm.IBM_MFM.IDAM])
                            idx = sec_data.find(idam)
                        else:
                            # Last sector: Look for GAP3
                            idx = sec_data.find(bytes([track.gapbyte]*16))
                        if idx > 0:
                            # 2 + gap_3 = CRC + GAP3 (because gap_included)
                            clippable += data_size - idx + 2 + gap_3
                            if bad_crc_clip_data:
                                data_size = idx
                                sec_data = sec_data[:data_size]
                                gap_included = True
                    elif data_size < native_size:
                        # Pad short data
                        sec_data += bytes(native_size - data_size)
                    elif data_size > native_size:
                        # Clip long data
                        if (sec_data[-13] != 0
                            and all([v==0 for v in sec_data[-12:]])):
                            # Includes next pre-sync: Clip it.
                            sec_data = sec_data[:-12]
                        # Long data includes CRC and GAP
                        gap_included = True
                    am = bytes([0xa1, 0xa1, 0xa1, dmark]) + sec_data
                    if gap_included:
                        t += mfm.encode(am[3:])
                    else:
                        crc = mfm.crc16.new(am).crcValue
                        if errs.data_crc_error:
                            crc ^= 0x5555
                        am += struct.pack('>H', crc)
                        t += mfm.encode(am[3:])
                        if sh:
                            # GAP3 for all but last sector
                            t += mfm.encode(bytes([track.gapbyte] * gap_3))
                            ngap3 += 1

                # Special 8K track handler
                ntrack = cls()._build_8k_track(sectors)
                if ntrack:
                    track = ntrack
                    break

                # The track may be too long to fit: Check for overhang.
                tracklen = int((track.time_per_rev / track.clock) / 16)
                overhang = int(len(t)//2 - tracklen*0.99)
                if overhang <= 0:
                    # We're done: Generate the pre-index gap
                    track.verify_len = len(t)*8
                    gap = tracklen - len(t)//2
                    t += mfm.encode(bytes([track.gapbyte] * gap))
                    track.bits = bitarray(endian='big')
                    track.bits.frombytes(mfm.mfm_encode(t))
                    break

                # Some EDSK tracks with Bad CRC contain a raw dump following
                # the DAM. This can usually be clipped.
                if clippable > overhang:
                    bad_crc_clip_data = True
                    continue

                # Some EDSK images have bogus GAP3 values. Shrink it if
                # necessary.
                new_gap_3 = -1
                if ngap3 != 0:
                    new_gap_3 = gap_3 - math.ceil(overhang / ngap3)
                error.check(new_gap_3 >= 0,
                            'EDSK: Track %d.%d is too long '
                            '(%d bits @ GAP3=%d; %d bits @ GAP3=0)'
                            % (cyl, head, len(t)*8, gap_3,
                               (len(t)//2-gap_3*ngap3)*16))
                #print('EDSK: GAP3 reduced (%d -> %d)' % (gap_3, new_gap_3))
                gap_3 = new_gap_3

            edsk.to_track[cyl,head] = track
            o += track_size

        return edsk


    def get_track(self, cyl, side):
        if (cyl,side) not in self.to_track:
            return None
        return self.to_track[cyl,side].raw_track()


# Local variables:
# python-indent: 4
# End: