No equivalent for ide-scsi available with the new PATA drivers?

No equivalent for ide-scsi available with the new PATA drivers?

[Chuck Ebbert]

People with IDE tape drives are complaining there's no ide-scsi
driver available anymore with the new PATA layer. Is anyone
working on one?

Re: No equivalent for ide-scsi available with the new PATA drivers?

[James Bottomley]

On Mon, 2007-07-23 at 12:36 -0400, Chuck Ebbert wrote:
> People with IDE tape drives are complaining there's no ide-scsi
> driver available anymore with the new PATA layer. Is anyone
> working on one?

You mean libata-pata?  The SCSI tape driver, st (or the onstream osst
driver), should just attach natively, shouldn't it ... or is there
something missing in the sat layer?

James

Re: No equivalent for ide-scsi available with the new PATA drivers?

[Alan Cox]

On Mon, 23 Jul 2007 12:36:44 -0400
Chuck Ebbert <cebbert@redhat.com> wrote:

> People with IDE tape drives are complaining there's no ide-scsi
> driver available anymore with the new PATA layer. Is anyone
> working on one?

It would make no sense. The libata layer is already using the SCSI
midlayer so compliant devices should already appear with the st driver.

We might need some tape detect logic, but I don't have any tape devices
so its up to someone with a tape drive handy to help out...

Alan

Re: No equivalent for ide-scsi available with the new PATA drivers?

[Chuck Ebbert]

On 07/23/2007 12:52 PM, Alan Cox wrote:
>> People with IDE tape drives are complaining there's no ide-scsi
>> driver available anymore with the new PATA layer. Is anyone
>> working on one?
> 
> It would make no sense. The libata layer is already using the SCSI
> midlayer so compliant devices should already appear with the st driver.
> 
> We might need some tape detect logic, but I don't have any tape devices
> so its up to someone with a tape drive handy to help out...

I should have posted this earlier:

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=243568

Re: No equivalent for ide-scsi available with the new PATA drivers?

[James Bottomley]

On Mon, 2007-07-23 at 13:01 -0400, Chuck Ebbert wrote:
> On 07/23/2007 12:52 PM, Alan Cox wrote:
> >> People with IDE tape drives are complaining there's no ide-scsi
> >> driver available anymore with the new PATA layer. Is anyone
> >> working on one?
> > 
> > It would make no sense. The libata layer is already using the SCSI
> > midlayer so compliant devices should already appear with the st driver.
> > 
> > We might need some tape detect logic, but I don't have any tape devices
> > so its up to someone with a tape drive handy to help out...
> 
> I should have posted this earlier:

It looks like an error handling issue ... which command was being
executed at the time the tape at ata2.01 went into the error state?

I haven't looked, but it's entirely possible that it's a simple timeout
issue:  libata does some of its own error handling, but the tape
commmands have to run with timeouts set to huge values (because it can
take so long to rewind a tape or write a long block).

James

Re: No equivalent for ide-scsi available with the new PATA drivers?

[Mark Lord]

Alan Cox wrote:
> On Mon, 23 Jul 2007 12:36:44 -0400
> Chuck Ebbert <cebbert@redhat.com> wrote:
> 
>> People with IDE tape drives are complaining there's no ide-scsi
>> driver available anymore with the new PATA layer. Is anyone
>> working on one?
> 
> It would make no sense. The libata layer is already using the SCSI
> midlayer so compliant devices should already appear with the st driver.
> 
> We might need some tape detect logic, but I don't have any tape devices
> so its up to someone with a tape drive handy to help out...

I have a PATA tape drive here somewhere.
Does anyone out there want it, with the understanding that it be
used to test/improve libata tape support?

One thing about ATAPI tape drives, is that they generally won't "stream"
with the generic SCSI st code.  Instead, many of them will do a *very* sluggish
stop/start stop/start stop/start kind of action when reading/writing bulk
data.  To get them to stream continously requires more driver intelligence,
*especially* when sharing a cable with some other device.

With an appropriate driver, tailored for ATAPI tape drives, they can run
at full speed while sharing a channel with another active ATA/ATAPI device,
both drives in use at the same time.

Cheers

Re: No equivalent for ide-scsi available with the new PATA drivers?

[Alan Cox]

> I have a PATA tape drive here somewhere.

-EBUSY

> Does anyone out there want it, with the understanding that it be
> used to test/improve libata tape support?

I'm happy to help someone else doing the work who needs info on libata
etc but I've no time to work on whats a very obscure corner case.