the media check problem
Dennis Bjorklund
db at zigo.dhs.org
Sun Nov 14 00:23:01 UTC 2004
On Sat, 13 Nov 2004, Alan Cox wrote:
> > 1) It's the read ahead that causes the problem. If one use
> > hdparm and set the read ahead to 0 on the device, then I have
> > no problem to verify a cd using dd.
>
> Correct
So, why not set read ahead to 0 during the install process (or at least
during the media check).
It's also strange that the media check works for some people.
I don't see anything here that depends on the hardware except perhaps
hardware that does not support dma at all (blacklisted), or that does not
use ide-cd. Maybe it's the combination of slow CPU and fast cd-rom that
makes it perform more read ahead and increase the chance of generating an
error (yet another guess :-).
> The "end" of a CD-R is undefined for about 150K. ide-scsi correctly
> retries errors at the disk end and reduces the disk size based on the
> error position. ide-cd doesn't.
If the layer above can tell ide-cd where the end is then it will never
need to seek past that. For an iso fs the size is known and I guessed in
the previous mail that this size was passed in to ide-cd when you mount
it.
A retry to find the real end of course would also solve it.
I don't know that much about cdroms so I can't say I know anything about
the "end is undefined for 150K" part. I played with different padding
sizes, but that didn't alter the end position. The full track I could read
back in was of constant size no matter what padding I tried.
Well, I feel a lot better now that I know what the problem is. I hate
problems where I don't know the technical reason behind it. Thanks.
--
/Dennis Björklund
More information about the fedora-devel-list
mailing list