[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: growisofs should have a method for padding

> The 2.6 kernel's CD/DVD driver has the unfortunate propensity to read
> ahead past the end of a CD/DVD.  It then reports an error, even if the
> actual read request was completely legitimate.

> The conventional work-around is to always pad when writing a CD/DVD.
> Ugly, but true.  (I agree that this isn't the correct fix, but the
> problem has not been fixed in a year,

Completely forget all your this-year's redhat references. This problem
existed at least with kernels 2.2 and 2.4, and is probably as old as
Linux (see references to it for kernels 1.3 in older mkisofs/cdrecord
source around the -pad option). In 2.2 it didn't usually show, in 2.4 it
was difficult to prevent. I've been pointing this out for years,
typically being ignored or looked at ("what's your problem?"). Part of
the reason may have been that the read-ahead I/O error doesn't show with
all CDs, and I've never observed it when reading all files on a CD with
kernels pre 2.6. Only when you read the whole ISO image do you run into
trouble with the last blocks. With 2.6 the read-ahead probably got
bigger and the I/O errors start earlier. Whether they reach into any
area occupied by files I don't know.

The net result is that you have to use padding somehow, and because
forgetting the option(s) produces a coaster I always use a wrapper
script for burning (it also makes me a log). With DVDs I haven't had any
errors recently, but can't say why.

Adding zeros to your image files doesn't change your checksums if you
compute them properly, ie on the ISO image only. I've got scripts for
all the above in my scriptutils package. It does damage any signatures
though, there you ought to use padding with mkisofs.

The default 20k padding was nowehre near enough on some DVD ISOs on 2.4.

I agree that it would be useful in growisofs. However a suitable wrapper
script should fix it too.

> 	/usr/bin/growisofs: no mkisofs options are permitted with =, aborting...

Well, if that "=" is with -Z, then growisofs is told to burn an image,
where you're already past mkisofs and therefore barfing on mkisofs
options is a good thing as it may save you a coaster (and makes no sense


Volker Kuhlmann			is possibly list0570 with the domain in header
http://volker.dnsalias.net/		Please do not CC list postings to me.

Reply to: