Re: [PATCH] fix req->cmd == INT cases

Re: [PATCH] fix req->cmd == INT cases

[Christoph Hellwig]

On Wed, Jun 20, 2007 at 01:53:32PM +0300, Boaz Harrosh wrote:
> Jens Axboe wrote:
> > 
> > If you look at most of the code, it's inside ifdef debug statements or
> > comments most of them. So I don't think you can base any removal
> > suggestion on that.
> > 
> > The patch itself looks fine though, if you send one that isn't mangled
> > I'll apply it to the 2.6.23 branch.
> > 
> Sorry new Thunderbird Installation. I forgot it does that.
> 
>  - I have unearthed very old bugs in stale drivers that still
>    used request->cmd as a READ|WRITE int
>  - This patch is maybe a proof that these drivers have not been
>    used for a long time. Should they be removed completely?
> 
> Drivers that currently do not work for sure:
>  drivers/acorn/block/fd1772.c
>  drivers/acorn/block/mfmhd.c

Afaik these are old arm26 driver, and that port is dead and should
probably be removed from the tree entirely.

>  drivers/cdrom/aztcd.c
>  drivers/cdrom/cm206.c
>  drivers/cdrom/gscd.c
>  drivers/cdrom/mcdx.c
>  drivers/cdrom/optcd.c
>  drivers/cdrom/sjcd.c

These are old cdrom drivers that are broken in various ways and probably
should be killed off aswell.

Re: [PATCH] fix req->cmd == INT cases

[Jens Axboe]

On Wed, Jun 20 2007, Christoph Hellwig wrote:
> On Wed, Jun 20, 2007 at 01:53:32PM +0300, Boaz Harrosh wrote:
> > Jens Axboe wrote:
> > > 
> > > If you look at most of the code, it's inside ifdef debug statements or
> > > comments most of them. So I don't think you can base any removal
> > > suggestion on that.
> > > 
> > > The patch itself looks fine though, if you send one that isn't mangled
> > > I'll apply it to the 2.6.23 branch.
> > > 
> > Sorry new Thunderbird Installation. I forgot it does that.
> > 
> >  - I have unearthed very old bugs in stale drivers that still
> >    used request->cmd as a READ|WRITE int
> >  - This patch is maybe a proof that these drivers have not been
> >    used for a long time. Should they be removed completely?
> > 
> > Drivers that currently do not work for sure:
> >  drivers/acorn/block/fd1772.c
> >  drivers/acorn/block/mfmhd.c
> 
> Afaik these are old arm26 driver, and that port is dead and should
> probably be removed from the tree entirely.
> 
> >  drivers/cdrom/aztcd.c
> >  drivers/cdrom/cm206.c
> >  drivers/cdrom/gscd.c
> >  drivers/cdrom/mcdx.c
> >  drivers/cdrom/optcd.c
> >  drivers/cdrom/sjcd.c
> 
> These are old cdrom drivers that are broken in various ways and probably
> should be killed off aswell.

I agree (with both sentiments), would anyone mind if I just killed them
off? mitsumi support is being reworked, that can get reintroduced once
the driver is in a stable state.

-- 
Jens Axboe

Re: [PATCH] fix req->cmd == INT cases

[Christoph Hellwig]

On Wed, Jun 20, 2007 at 02:24:09PM +0200, Jens Axboe wrote:
> > These are old cdrom drivers that are broken in various ways and probably
> > should be killed off aswell.
> 
> I agree (with both sentiments), would anyone mind if I just killed them
> off? mitsumi support is being reworked, that can get reintroduced once
> the driver is in a stable state.

Yeah - the new mitsumi doesn't share anything with the old one anyway.

Re: [PATCH] fix req->cmd == INT cases

[Rene Herman]

On 06/20/2007 02:24 PM, Jens Axboe wrote:

> On Wed, Jun 20 2007, Christoph Hellwig wrote:
>> On Wed, Jun 20, 2007 at 01:53:32PM +0300, Boaz Harrosh wrote:

>>>  drivers/cdrom/aztcd.c
>>>  drivers/cdrom/cm206.c
>>>  drivers/cdrom/gscd.c
>>>  drivers/cdrom/mcdx.c
>>>  drivers/cdrom/optcd.c
>>>  drivers/cdrom/sjcd.c
>>> 
>> These are old cdrom drivers that are broken in various ways and
>> probably should be killed off aswell.
> 
> I agree (with both sentiments), would anyone mind if I just killed them
> off?

I wouldn't. As far as I'm concerned everything in drivers/cdrom except 
viocd.c and cdrom.c itself can go, assuming I'll be allowed to bring back 
support for the legacy cdrom types I'd still like to have supported -- in 
the first place mitsumi (mcdx), panasonic (sbpcd) and sony (cdu31a) and 
perhaps at some point other types if/when I happen across the hardware.

The old drivers serve as a source of hardware information but at least mcdx 
was broken in so many ways that it only did so at the source level. When I 
wanted to check throughput with the old driver I actually had to go back as 
far as 2.0.34 to find a working driver (the old mcd.c, already removed since 
2.6.10). 2.0.34 sources are available from kernel.org, and 2.6.22 sources 
even from my local machine...

> mitsumi support is being reworked, that can get reintroduced once the
> driver is in a stable state.

... which, by the way, is still waiting on comment from anyone with a clue 
as to why it makes the machine go boom (easily repeatable when using CFQ, 
not or not easily when using AS):

http://lkml.org/lkml/2007/6/4/50

Rene.

Re: [PATCH] fix req->cmd == INT cases

[Jens Axboe]

On Wed, Jun 20 2007, Rene Herman wrote:
> On 06/20/2007 02:24 PM, Jens Axboe wrote:
> 
> >On Wed, Jun 20 2007, Christoph Hellwig wrote:
> >>On Wed, Jun 20, 2007 at 01:53:32PM +0300, Boaz Harrosh wrote:
> 
> >>> drivers/cdrom/aztcd.c
> >>> drivers/cdrom/cm206.c
> >>> drivers/cdrom/gscd.c
> >>> drivers/cdrom/mcdx.c
> >>> drivers/cdrom/optcd.c
> >>> drivers/cdrom/sjcd.c
> >>>
> >>These are old cdrom drivers that are broken in various ways and
> >>probably should be killed off aswell.
> >
> >I agree (with both sentiments), would anyone mind if I just killed them
> >off?
> 
> I wouldn't. As far as I'm concerned everything in drivers/cdrom except 
> viocd.c and cdrom.c itself can go, assuming I'll be allowed to bring back 
> support for the legacy cdrom types I'd still like to have supported -- in 
> the first place mitsumi (mcdx), panasonic (sbpcd) and sony (cdu31a) and 
> perhaps at some point other types if/when I happen across the hardware.

Sure, a working clean driver would always be allowed in. I'll remove the
legacy drivers in 2.6.23.

> The old drivers serve as a source of hardware information but at least
> mcdx was broken in so many ways that it only did so at the source
> level. When I wanted to check throughput with the old driver I
> actually had to go back as far as 2.0.34 to find a working driver (the
> old mcd.c, already removed since 2.6.10). 2.0.34 sources are available
> from kernel.org, and 2.6.22 sources even from my local machine...

Heh, that's pretty scary!

> >mitsumi support is being reworked, that can get reintroduced once the
> >driver is in a stable state.
> 
> ... which, by the way, is still waiting on comment from anyone with a clue 
> as to why it makes the machine go boom (easily repeatable when using CFQ, 
> not or not easily when using AS):
> 
> http://lkml.org/lkml/2007/6/4/50

Yeah I know, I will get around to it eventually.

-- 
Jens Axboe