Bug#319457: kernel-image-2.6.8-2-k7: DVD Burning with ide-scsi : oops the kernel, hangs ide-scsi
On Tue, Jul 26, 2005 at 10:44:49AM +0200, A Mennucc wrote:
> (I am sending this to cdrecord / dvdrecord / gcombust mantainers
> as well: please read below)
> On Mon, Jul 25, 2005 at 06:30:25PM +0900, Horms wrote:
> > tag 319457 +wontfix
> > thanks
> > On Fri, Jul 22, 2005 at 10:38:47AM +0200, A Mennucc wrote:
> > > Package: kernel-image-2.6.8-2-k7
> > > Version: 2.6.8-16
> > > Severity: important
> > >
> > > I just bought a DVD Burner; I wanted to burn stuff, so I loaded
> > > ide-scsi (since this was the way to go some years ago - I did not
> > > know that now is deprecated) ; the command
> > > $ dvdrecord dev=1,0,0 -v -dummy -dao debian-31r0a-i386-binary-1.iso
> > > segfaulted, and its fork started printing:
> > > dvdrecord: No such device or address. Cannot send SCSI cmd via ioctl
> > >
> > > Moreover the kernel oopsed (NULL pointer dereference)
> > > and then the module ide-scsi seems dead
> > > (cannot remove it, does not respond); see dmesg in attachment
> > Don't use ide-scsi, ever!
> two options:
> 1) If idescsi should never be used, then idescsi should
> not be built into the kernel (by setting BLK_DEV_IDESCSI to no)
Yes, I agree entirely. However, I was asked to leave it in
so as not to break anything that happens to work using it.
So its there for 2.6.8.
> 2) If idescsi is shipped, then it should not ever oops by
> NULL point dereferencing.
> Your tag 'wontfix' is nonsense: any option of the above that you
> choose, you find a a bug in the current kernel that must be fixed in
> future revisions.
I disagree. I think there is some value in leaving this driver in
as it does work for some people. Think of it as a transition process.
In any case, its there for sarge, so its too late. I certainly would
like to see it disabled for the Etch kernel.
> > Can dvdrecord use ide devices directly?
> When starting cdrecord using ATAPI interface it prints
> Warning: Using ATA Packet interface.
> Warning: The related Linux kernel interface code seems to be unmaintained.
> Warning: There is absolutely NO DMA, operations thus are slow.
> This seems to suggest that idescsi should be used instead.
> Currently /usr/share/doc/cdrecord/README.ATAPI.setup says
> "Usually you would not want to use ide-scsi with kernel 2.6"
> It does not say : "Do not ever use it"
> Moreover gcombust does not support ATAPI in its interface
> > Any tool that uses it should have a bug filed against it.
> > If so please do that.
> > I'm marking this as wontfix and leaving it open
> > so it can act as a reference for others.
> I think that kernel people and the mantainer of cdrecord/dvdrecord
> should reach an agreement on this issue.
> (Translation: I am just a poor user, please sort this out between
Yes, I agree.
> If idescsi should not be used, then the documents of
> cdrecord/dvdrecord should be adjusted. Moreover gcombust needs some
> refurbishment to be able to use ATAPI.
Yes, I agree.