ide_cd modifies hard_* members of request

ide_cd modifies hard_* members of request

[Nikanth Karthikesan]

Hi

Despite the warning in blkdev.h, that members of struct request starting
with hard_* are block layer internals, and no driver should touch them,
the ide_cd driver seems to fiddle around with it.

Is this reqd, or can this be made to work without the need for this?

If this reqd, why not provide a helper function for that in block layer
itself, may be, with a warning about usage?

Thanks
Nikanth Karthikesan

Re: ide_cd modifies hard_* members of request

[Boris Petkov]

(sorry, forgot to cc all)

On Tue, Mar 18, 2008 at 9:54 AM, Nikanth Karthikesan <knikanth@suse.de> wrote:
> Hi
>
>  Despite the warning in blkdev.h, that members of struct request starting
>  with hard_* are block layer internals, and no driver should touch them,
>  the ide_cd driver seems to fiddle around with it.
>
>  Is this reqd, or can this be made to work without the need for this?
>
>  If this reqd, why not provide a helper function for that in block layer
>  itself, may be, with a warning about usage?
>
>  Thanks
>  Nikanth Karthikesan

Hi,

the code pieces you're referring to are in there since at least the
first git import of the kernel source (i.e. 2005) aka long time so
Bart/Jens should know that...


-- 
Regards/Gruß,
Boris

Re: ide_cd modifies hard_* members of request

[Jens Axboe]

On Tue, Mar 18 2008, Boris Petkov wrote:
> (sorry, forgot to cc all)
> 
> On Tue, Mar 18, 2008 at 9:54 AM, Nikanth Karthikesan <knikanth@suse.de> wrote:
> > Hi
> >
> >  Despite the warning in blkdev.h, that members of struct request starting
> >  with hard_* are block layer internals, and no driver should touch them,
> >  the ide_cd driver seems to fiddle around with it.
> >
> >  Is this reqd, or can this be made to work without the need for this?
> >
> >  If this reqd, why not provide a helper function for that in block layer
> >  itself, may be, with a warning about usage?
> >
> >  Thanks
> >  Nikanth Karthikesan
> 
> Hi,
> 
> the code pieces you're referring to are in there since at least the
> first git import of the kernel source (i.e. 2005) aka long time so
> Bart/Jens should know that...

restore_request(), I'm assuming. It was probably me who modified this
code, I don't think there's any justification for it. I would suggest
killing it. And cdrom_read_from_buffer() as well, it's a silly addition.

-- 
Jens Axboe

Re: ide_cd modifies hard_* members of request

[Boris Petkov]

On Tue, Mar 18, 2008 at 12:08 PM, Jens Axboe <jens.axboe@oracle.com> wrote:
>
> On Tue, Mar 18 2008, Boris Petkov wrote:
>  > (sorry, forgot to cc all)
>  >
>  > On Tue, Mar 18, 2008 at 9:54 AM, Nikanth Karthikesan <knikanth@suse.de> wrote:
>  > > Hi
>  > >
>  > >  Despite the warning in blkdev.h, that members of struct request starting
>  > >  with hard_* are block layer internals, and no driver should touch them,
>  > >  the ide_cd driver seems to fiddle around with it.
>  > >
>  > >  Is this reqd, or can this be made to work without the need for this?
>  > >
>  > >  If this reqd, why not provide a helper function for that in block layer
>  > >  itself, may be, with a warning about usage?
>  > >
>  > >  Thanks
>  > >  Nikanth Karthikesan
>  >
>  > Hi,
>  >
>  > the code pieces you're referring to are in there since at least the
>  > first git import of the kernel source (i.e. 2005) aka long time so
>  > Bart/Jens should know that...
>
>  restore_request(), I'm assuming. It was probably me who modified this
>  code, I don't think there's any justification for it. I would suggest
>  killing it.

I'll conjure up something later wrt above.

> And cdrom_read_from_buffer() as well, it's a silly addition.

This one is already gone, however the patch is still in Bart's tree
for -mm testing:
ftp://ftp.fi.kernel.org/pub/linux/kernel/people/bart/pata-2.6/patches/ide-cd-remove-the-internal-64k-buffer.patch

>
>  --
>  Jens Axboe
>
>



-- 
Regards/Gruß,
Boris