Peter Jones wrote:
On Wed, 2006-03-29 at 05:49 -0500, Alan Cox wrote:
> On Tue, Mar 28, 2006 at 02:31:19PM -0800, Schlaegel wrote:
>> This presents the obvious question: why don't we just check the iso on
>> the media instead of the media itself, and thus remove this long
>> standing problem.
> If we wanted to do that we would need to do it by walking the mounted iso
> file system. The moment we deal with the raw device the kernel starts treating
> it as a device not a file system so we get readehead into the twilight zone.
>
> However I'm glad you asked because you've made me realise we may have the
bits
> to deal with this as of FC4/5. Specifically we could use the device mapper to
> create an ISO sized mapping of the relevant chunk of the CD and check that...
>
> That might just work
I'm certainly willing to try this...
... but we still have a lingering problem with dm where *it* tries to
read off the end of its underlying devices.
But seriously, somebody tell me a drive make and model, media
manufacturer, cdrecord (or other tool) command line, and FC5 disc number
that reliably has this failure, and I'll be happy to try this or any
other idea people have to fix it.
This may or may not be related but it has the same(ide=nodma) workaround:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187203
entered last night. The h/w is pretty generic. If you need more details
of the h/w, let me know. the media is Memorex CD-RW 700MB.
Richard