Re: dvd+rw-tools-5.20.4.10.8: File is too large
me > > ... the annoying read-ahead bug ...
> > data format but to CD reading on Linux. (I do not experience
> > it with growisofs written DVD+RW on Linux)
Volker Kuhlmann >
> That would probably be because the dvd+rw has previously been written
> to beyond the current end of the recording. I.e. the read-ahead does
> find readable sectors.
Actually not with my drives. Since i had reason to log my first
experiences with the writer in june i can quite surely state that my
DVD-ROM stops reading at the end of the written area, delivering
the right amount of bytes (including 128k of selfmade padding).
The DVD writer does read 4.7e9 bytes even if only 1 GB has ever been
written.
I tested 20 DVD+RW media with 4480 MB (1 heavily ill already when writing).
Then i tried to find out about end-of-file detection and tested again
8 times with 4482 MB. I carefully looked wether the DVDs returned
enough data with both drives.
So at least with my equipment, there is no read-ahead-bug behavior.
Nevertheless, the differences between both drives make me think.
(one is /dev/hdd, the other /dev/sr0)
I dug in my project's mail archive and found 14 users who report
about DVD writing and subsequent verifying (mostly DVD+RW, some
DVD+R, some undetermined). None of them reports a problem about
missing bytes at the end. The verification run would have noticed
if more than the 128k of padding had been missing.
> ... an asus E616 dvdrom drive produces one IDE bus error
> after the other, ...
Uh oh. A poisonous backup DVD. Quite a nightmare.
> > because mkisofs adds a generous amount of padding by default.
> > (300 kB according to 2.01a35's man mkisofs, option "-pad")
>
> It used to be 30k, which was enough for 2.2 but not for 2.4 kernels. I
> found that adding >0.5MB was enough in the cases I tested, though use
> 1MB.
Your report makes me ponder wether the 300 kB of mkisofs are still
safe. I add my own 128k but ... uh oh, again.
I dug in my memory and found out why i believed to be safe from
read-ahaed bugs. It was 14 Mar 2003.
I told Andy Polyakov about the drive of my first DVD user :
>> ... drive was still not willing to read the last 73 kByte and gnawed
>> on the disk for nearly 10 minutes until it finally gave up.
Andy asked back :
> Can you confirm that it's DVD+RW media and not DVD+R? If DVD+RW, was
> lead-out explicitly written with either 'growisofs -dvd-compat ...' or
> 'dvd+rw-format -lead-out /dev/scd0'? A.
Finally the problem turned out to be caused by outdated kernel
buffer content. I did not reload the DVD before checkreading it.
(growisofs now takes care of this)
But Andy seemed to expect that this pseudo read-ahead bug
was caused by me not using -dvd-compat . (I didn't, but do now)
It might be that i confused that and thought to be safe just
because no read problems did happen any more.
Therefore a question to Andy :
Is there a relation between -dvd-compat with DVD+RW and not
experiencing missing bytes at the end of the written part
of the media ?
(I understand that -dvd-compat puts the leadout at the end of the
data area. But does that mean anything for readability of the
data's tail by the common Linux drivers ?)
Have a nice day :)
Thomas
Reply to: