Partially complete rework of all the decoders to avoid seeking inside the

fluxmap. This requires resetting the FluxDecoder, which loses any pending
state, resulting in bad reads for (some) formats which don't have gaps between
sectors --- the DVK MX is the main victim.
This commit is contained in:
David Given
2022-02-12 00:55:09 +01:00
parent 610b7fe95c
commit 0933dc1afa
20 changed files with 204 additions and 227 deletions

View File

@@ -98,58 +98,55 @@ public:
_config(config.ibm())
{}
RecordType advanceToNextRecord() override
nanoseconds_t advanceToNextRecord() override
{
const FluxMatcher* matcher = nullptr;
_sector->clock = _fmr->seekToPattern(ANY_RECORD_PATTERN, matcher);
/* If this is the MFM prefix byte, the the decoder is going to expect three
* extra bytes on the front of the header. */
_currentHeaderLength = (matcher == &MFM_PATTERN) ? 3 : 0;
Fluxmap::Position here = tell();
resetFluxDecoder();
if (_currentHeaderLength > 0)
readRawBits(_currentHeaderLength*16);
auto idbits = readRawBits(16);
const Bytes idbytes = decodeFmMfm(idbits);
uint8_t id = idbytes.slice(0, 1)[0];
if (eof())
return RecordType::UNKNOWN_RECORD;
seek(here);
switch (id)
{
case IBM_IDAM:
return RecordType::SECTOR_RECORD;
case IBM_DAM1:
case IBM_DAM2:
case IBM_TRS80DAM1:
case IBM_TRS80DAM2:
return RecordType::DATA_RECORD;
}
return RecordType::UNKNOWN_RECORD;
return seekToPattern(ANY_RECORD_PATTERN);
}
void decodeSectorRecord() override
{
unsigned recordSize = _currentHeaderLength + IBM_IDAM_LEN;
auto bits = readRawBits(recordSize*16);
auto bytes = decodeFmMfm(bits).slice(0, recordSize);
/* This is really annoying because the IBM record scheme has a
* variable-sized header _and_ the checksum covers this header too. So
* we have to read and decode a byte at a time until we know where the
* record itself starts, saving the bytes for the checksumming later.
*/
Bytes bytes;
ByteWriter bw(bytes);
auto readByte = [&]() {
auto bits = readRawBits(16);
auto bytes = decodeFmMfm(bits).slice(0, 1);
uint8_t byte = bytes[0];
bw.write_8(byte);
return byte;
};
uint8_t id = readByte();
if (id == 0xa1)
{
readByte();
readByte();
id = readByte();
}
if (id != IBM_IDAM)
return;
ByteReader br(bytes);
br.seek(bw.pos);
auto bits = readRawBits(IBM_IDAM_LEN*16);
bw += decodeFmMfm(bits).slice(0, IBM_IDAM_LEN);
IbmDecoderProto::TrackdataProto trackdata;
getTrackFormat(trackdata, _sector->physicalCylinder, _sector->physicalHead);
ByteReader br(bytes);
br.seek(_currentHeaderLength);
br.read_8(); /* skip ID byte */
_sector->logicalTrack = br.read_8();
_sector->logicalSide = br.read_8();
_sector->logicalSector = br.read_8();
_currentSectorSize = 1 << (br.read_8() + 7);
uint16_t gotCrc = crc16(CCITT_POLY, bytes.slice(0, br.pos));
uint16_t wantCrc = br.read_be16();
uint16_t gotCrc = crc16(CCITT_POLY, bytes.slice(0, _currentHeaderLength + 5));
if (wantCrc == gotCrc)
_sector->status = Sector::DATA_MISSING; /* correct but unintuitive */
@@ -163,17 +160,39 @@ public:
void decodeDataRecord() override
{
unsigned recordLength = _currentHeaderLength + _currentSectorSize + 3;
auto bits = readRawBits(recordLength*16);
auto bytes = decodeFmMfm(bits).slice(0, recordLength);
/* This is the same deal as the sector record. */
Bytes bytes;
ByteWriter bw(bytes);
auto readByte = [&]() {
auto bits = readRawBits(16);
auto bytes = decodeFmMfm(bits).slice(0, 1);
uint8_t byte = bytes[0];
bw.write_8(byte);
return byte;
};
uint8_t id = readByte();
if (id == 0xa1)
{
readByte();
readByte();
id = readByte();
}
if ((id != IBM_DAM1) && (id != IBM_DAM2)
&& (id != IBM_TRS80DAM1) && (id != IBM_TRS80DAM2))
return;
ByteReader br(bytes);
br.seek(_currentHeaderLength);
br.read_8(); /* skip ID byte */
br.seek(bw.pos);
auto bits = readRawBits((_currentSectorSize + 2) * 16);
bw += decodeFmMfm(bits).slice(0, _currentSectorSize+2);
_sector->data = br.read(_currentSectorSize);
uint16_t gotCrc = crc16(CCITT_POLY, bytes.slice(0, br.pos));
uint16_t wantCrc = br.read_be16();
uint16_t gotCrc = crc16(CCITT_POLY, bytes.slice(0, recordLength-2));
_sector->status = (wantCrc == gotCrc) ? Sector::OK : Sector::BAD_CHECKSUM;
}