CD writing in future Linux try #2 [ was: Re: CD writing in future Linux (stirring up a hornets' nest) ]

[Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>]

Hi,

I'm stealing thread in hope of stopping flamewar
and finally having some useful discussion.

I address this to everybody not only to Joerg:

[*] As this is my thread now discussing SG_IO on /dev/hd* vs /dev/sg*
is BANNED.  Please continue discussing this in the old thread. :-)

I'm sure thare are other important things that we can agree on.

If it doesn't work it will be my first and the last mail in this thread...

On 1/27/06, Joerg Schilling <schilling@fokus.fraunhofer.de> wrote:
> Vojtech Pavlik <vojtech@suse.cz> wrote:
>
> > > The only integrative (and this useful for libscg) interface on Linux is /dev/sg*
> > >
> > > /dev/hd* may look nice if you only look skin-deep
> > >
> > > How do you e.g. like send SCSI commands to ATAPI tape drives on Linux?
> >
> > I see you asking this again and again, and you seem to want to hear this
> > answer: "You don't." I haven't checked the code, but I guess the SG_IO
> > interface isn't available there. And I don't think this is a problem,
> > because a) if it was needed, it can be added trivially with minimum
> > added code, b) ATAPI tapes are dead, much the way ATAPI floppies are.
>
> Nice to see that agree that sending SCSI via /dev/hd* is a hack with limited
> benefit.
>
> Maybe (now that we did agree about the way to go) it makes sense to start
> discussing which bugs in Linux need to be fixed in order to close e.g.
> the Bugs in the Debian bug tracking system for cdrtools that are related to the
> Linux kernel.

Yes, lets talk about other problems, do you have pointers bugs handy?
I'm not very familiar with Debian bug tracking system and I see there
more than 3 bugs for cdrtools.

> This are the three most important Linux kernel bugs:
>
> -       ide-scsi does not do DMA if the DMAsize is not a multiple of 512
>         A person who knows internal Linux structures shoule be able
>         to fix this (and allow any multiple of 4) in less than one hour.

I'll take a look, it should be quite easy
and I don't see a reason for this restriction.

>         If we sum up the time spend on this discussoion, I really cannot
>         understand why this has not been fixed earlier.

Because nobody cared and flamewaring is easy in contrast
to doing some real work.

> -       /dev/hd* artificially prevents the ioctls SCSI_IOCTL_GET_IDLUN
>         SCSI_IOCTL_GET_BUS_NUMBER from returning useful values.
>         As long as this bug is present, there is no way to see SG_IO
>         via /dev/hd* as integral part of the Linux SCSI transport concept.

What are the return values of SCSI_IOCTL_GET_IDLUN
and SCSI_IOCTL_GET_BUS_NUMBER needed for?

Please elaborate as:

* SG_IO ioctl doesn't require lun and bus number for arguments
* sg_io_hdr_t also doesn't contain/require these fields

so I simply cannot see why they are needed by kernel.

If lun and bus number are needed by cdrecord please explain why
(if only for enumerating devices sorry but see [*]).

> -       If sending SCSI sia ATAPI, Linux under certain unknown conditions
>         bastardizes the content of SCSI commands. A possible reason may be

Sorry but I can't do much about it without knowing more details.

Please provide some way to reproduce the problem
("certain unknown conditions" is not very useful).

>         that it sends random data in the remainder between the actual
>         SCSI cdb size and the ATAPI SCSI cdb size.

It should send 0-s but I'll recheck this.

>         ATAPI drives that verify SCSI cdb's for correctness fail in this
>         case.

Thanks,
Bartlomiej