Re: ide-tape broken (was Re: [PATCH] use ide-identify.h, fix endian bug)

Re: ide-tape broken (was Re: [PATCH] use ide-identify.h, fix endian bug)

[Mikael Pettersson]

On Thu, 7 Aug 2003 22:51:31 +0200 (MET DST), Bartlomiej Zolnierkiewicz wrote:
>On Thu, 7 Aug 2003 Andries.Brouwer@cwi.nl wrote:
>
>> Given ide-identify.h, we can simplify ide-floppy.c and ide-tape.c a lot.
>> In fact ide-tape.c was broken on big-endian machines.
>> (Unfortunately much more is broken that was fixed here,
>> ide-tape.c is not in a good shape today.)
>
>ide-tape is broken because nobody cares, so I don't care too
>(was broken even before).  It needs rewrite and testing.
>
>So once again if anybody cares and has hardware to test,
>please contact me and I will try fix it.
>
>Until then I don't touch it et all and consider it obsoleted.

I used to use the ide-tape driver for my Seagate STT8000A,
but its prolonged brokenness in 2.5 (caused initially by the
bio changes) made me switch to ide-scsi + st instead since
that at least _works_.

I can test updates as they hit Linus 2.6-releases, but frankly
I'd rather use ide-scsi+st or a new clean ATAPI tape driver
than ide-tape.c. (I've studied ide-tape.c. It reeks of poor
coding style, kludges for Onstream, and an over-engineered
buffering scheme. And it's known to have problems with DMA.)

/Mikael

Re: ide-tape broken (was Re: [PATCH] use ide-identify.h, fix endian bug)

[Bartlomiej Zolnierkiewicz]

On Fri, 8 Aug 2003, Mikael Pettersson wrote:

> On Thu, 7 Aug 2003 22:51:31 +0200 (MET DST), Bartlomiej Zolnierkiewicz wrote:
> >On Thu, 7 Aug 2003 Andries.Brouwer@cwi.nl wrote:
> >
> >> Given ide-identify.h, we can simplify ide-floppy.c and ide-tape.c a lot.
> >> In fact ide-tape.c was broken on big-endian machines.
> >> (Unfortunately much more is broken that was fixed here,
> >> ide-tape.c is not in a good shape today.)
> >
> >ide-tape is broken because nobody cares, so I don't care too
> >(was broken even before).  It needs rewrite and testing.
> >
> >So once again if anybody cares and has hardware to test,
> >please contact me and I will try fix it.
> >
> >Until then I don't touch it et all and consider it obsoleted.
>
> I used to use the ide-tape driver for my Seagate STT8000A,
> but its prolonged brokenness in 2.5 (caused initially by the
> bio changes) made me switch to ide-scsi + st instead since
> that at least _works_.
>
> I can test updates as they hit Linus 2.6-releases, but frankly

Great.

> I'd rather use ide-scsi+st or a new clean ATAPI tape driver
> than ide-tape.c. (I've studied ide-tape.c. It reeks of poor

"new clean ATAPI tape drive" == the one yet to be written?

> coding style, kludges for Onstream, and an over-engineered
> buffering scheme. And it's known to have problems with DMA.)

So you are familiar with the code, cool. ;-)

I will try to simplify ide-tape.c as much as possible.

--bartlomiej

Re: ide-tape broken (was Re: [PATCH] use ide-identify.h, fix endian bug)

[Alan Cox]

On Gwe, 2003-08-08 at 01:10, Bartlomiej Zolnierkiewicz wrote:
> "new clean ATAPI tape drive" == the one yet to be written?
> 
> > coding style, kludges for Onstream, and an over-engineered
> > buffering scheme. And it's known to have problems with DMA.)
> 
> So you are familiar with the code, cool. ;-)
> 
> I will try to simplify ide-tape.c as much as possible.

A starting point might actually be cp ide-tape.c ide-tape-ontrack.c and
then zap the conditional mess in both. Thats roughly how scsi handles
it all with st and osst

Re: ide-tape broken (was Re: [PATCH] use ide-identify.h, fix endian bug)

[Mikael Pettersson]

On Fri, 8 Aug 2003 02:10:19 +0200 (MET DST), Bartlomiej Zolnierkiewicz wrote:
>> I'd rather use ide-scsi+st or a new clean ATAPI tape driver
>> than ide-tape.c. (I've studied ide-tape.c. It reeks of poor
>
>"new clean ATAPI tape drive" == the one yet to be written?

Correct.

>> coding style, kludges for Onstream, and an over-engineered
>> buffering scheme. And it's known to have problems with DMA.)
>
>So you are familiar with the code, cool. ;-)

Only as a generic coder trying to fix it, and as a user
looking at other people's bug reports.

/Mikael

ide-tape broken (was Re: [PATCH] use ide-identify.h, fix endian bug)

[Bartlomiej Zolnierkiewicz]

On Thu, 7 Aug 2003 Andries.Brouwer@cwi.nl wrote:

> Given ide-identify.h, we can simplify ide-floppy.c and ide-tape.c a lot.
> In fact ide-tape.c was broken on big-endian machines.
> (Unfortunately much more is broken that was fixed here,
> ide-tape.c is not in a good shape today.)

ide-tape is broken because nobody cares, so I don't care too
(was broken even before).  It needs rewrite and testing.

So once again if anybody cares and has hardware to test,
please contact me and I will try fix it.

Until then I don't touch it et all and consider it obsoleted.

--bartlomiej