[Wylug-help] V-CD problem

[John Hodrien]

Aaron Crane wrote:
> John Hodrien writes:
>> The fact he's getting I/O errors suggests there's something wrong when 
>> linux mounts the disk.  Normally you'd assume this was a faulty disk, but 
>> seeing as it works fine in XP, I'm curious.
> 
> Marginal media?  Conceivably the poster's combination of (IDE chipset,
> drive) works better with dodgy disks under the XP driver than the Linux
> one.  But, yeah, the I/O errors do make it look from here like there's
> something wrong with the disk.
> 
>>> Googling for avseq##.dat provides several contradictory accounts.
>>> Some suggest that avseq##.dat are pointers/shortcuts to the actual
>>> files rather than the data files themselves. 
>> I've no idea, but if he's copied the .dat file off in XP and it plays
>> under linux, I guess not.  I was under the impression that they're
>> just normal iso9660 CDs but with a specific directory structure.
> 
> VCDs do have an ISO-9660 filesystem, and the AVSEQ##.DAT files do appear
> in it.  However, the data blocks in those files normally appear in a
> mode 2 (or mode 2 form 2) track on the CD, rather than in the mode 1
> track that contains the filesystem metadata and the data for the other
> files.

I was kinda hedging that there was something like this going on.  If that was 
the case, and the media is perfect, could it be that we're just treating it like 
a iso9660, and getting confused by the error correction not matching the data. 
Make sure the CD is not mounted, and try xine vcd:/

> What's the difference?  Well, mode 1 tracks have 2048-byte sectors,
> which works well for ordinary data.  Except that's a slight
> simplification; each sector is actually 2048 bytes plus 304 bytes of
> error correction.  Mode 1 sectors use all the error-correction bytes
> for, well, error correction, on the assumption that a sector error in
> the middle of arbitrary data would be bad.  Audio CD tracks use all 304
> bytes as additional data, so they have 2352 bytes per sector (588
> samples, or 1/75 seconds).  Mode 2 tracks use 16 bytes for error
> correction, giving 2336 bytes per sector, and mode 2 form 2 tracks use
> 28 bytes for error correction, giving 2324 bytes per sector.
> 
> This makes some sense: human perception of audio and video data is
> fairly robust, so error correction isn't as important as for other
> things.  The tradeoff between reliability and storage capacity is made
> in a different place.

I didn't know the specifics, but knew Audio CDs did that.  Thanks for the detail.

> So, with regard to reading a VCD as an ISO-9660 filesystem: the files
> are there, but when the kernel reads the AVSEQ##.DAT files, it only
> looks at the 2048 data bytes in each sector, so the file you'd get is
> broken.  For VCDs that use mode 2 tracks, you need to access the video
> data without going through the filesystem.

And results in an Input/Output error d'ya think?

ja