ioctl32: kill code duplication (sparc64 tester wanted)

ioctl32: kill code duplication (sparc64 tester wanted)

[Pavel Machek]

[-- Attachment #1: Type: text/plain, Size: 267 bytes --]

Hi!

Attached patch shares ioctl32 handles between x86-64 and sparc64 (and
it should be possible/easy to share with other archs, too).

I'd like sparc64 person to test/comment on it...
						Pavel
-- 
Horseback riding is like software...
...vgf orggre jura vgf serr.

[-- Attachment #2: delme.gz --]
[-- Type: application/octet-stream, Size: 50557 bytes --]

Re: ioctl32: kill code duplication (sparc64 tester wanted)

[Christoph Hellwig]

On Mon, May 12, 2003 at 01:40:55PM +0200, Pavel Machek wrote:
> Hi!
> 
> Attached patch shares ioctl32 handles between x86-64 and sparc64 (and
> it should be possible/easy to share with other archs, too).
> 
> I'd like sparc64 person to test/comment on it...

I don't have a sparc64, but there's certainly no <asm/mtrr.h> for that arch..

Also #including c files is ugly as hell.  What's the #ifdef INCLUDES
supposed to help?

Re: ioctl32: kill code duplication (sparc64 tester wanted)

[Pavel Machek]

Hi!

> > Attached patch shares ioctl32 handles between x86-64 and sparc64 (and
> > it should be possible/easy to share with other archs, too).
> > 
> > I'd like sparc64 person to test/comment on it...
> 
> I don't have a sparc64, but there's certainly no <asm/mtrr.h> for
> > that arch..

I thought I killed that one?

> Also #including c files is ugly as hell.  What's the #ifdef INCLUDES
> supposed to help?

Yes, but do you have better proposal how to kill 4000+ lines of code
from each 64-bit architecture?
								Pavel
-- 
Horseback riding is like software...
...vgf orggre jura vgf serr.

Re: ioctl32: kill code duplication (sparc64 tester wanted)

[Christoph Hellwig]

On Mon, May 12, 2003 at 03:05:19PM +0200, Pavel Machek wrote:
> Hi!
> > I don't have a sparc64, but there's certainly no <asm/mtrr.h> for
> > > that arch..
> 
> I thought I killed that one?

The patch you attached added it..

> > Also #including c files is ugly as hell.  What's the #ifdef INCLUDES
> > supposed to help?
> 
> Yes, but do you have better proposal how to kill 4000+ lines of code
> from each 64-bit architecture?

What's the reason you can't build fs/compat_ioctl.c normally and pull
in the arch magic through a magic asm/ header?  You still haven't answered
the second question, btw.. 

Re: ioctl32: kill code duplication (sparc64 tester wanted)

[Pavel Machek]

Hi!

> > > I don't have a sparc64, but there's certainly no <asm/mtrr.h> for
> > > > that arch..
> > 
> > I thought I killed that one?
> 
> The patch you attached added it..

There are probably more such mistake. It should be easy for anyone
with sparc64 to kill them all at once.

> > > Also #including c files is ugly as hell.  What's the #ifdef INCLUDES
> > > supposed to help?
> > 
> > Yes, but do you have better proposal how to kill 4000+ lines of code
> > from each 64-bit architecture?
> 
> What's the reason you can't build fs/compat_ioctl.c normally and pull
> in the arch magic through a magic asm/ header?  

Some architectures need special stuff (mtrr's), so I'd have to include
.c files, too (the other way). [Look at how the table of ioctls is
generated, its asm magic].

> You still haven't answered
> the second question, btw.. 

Are you asking why are there #includes in compat_ioctl.c? Its because
there is so many of them, and having to update all archs when you
tuoch fs/compat_ioctl.c would be bad.

								Pavel
-- 
Horseback riding is like software...
...vgf orggre jura vgf serr.

Re: ioctl32: kill code duplication (sparc64 tester wanted)

[Christoph Hellwig]

On Mon, May 12, 2003 at 03:13:26PM +0200, Pavel Machek wrote:
> > What's the reason you can't build fs/compat_ioctl.c normally and pull
> > in the arch magic through a magic asm/ header?  
> 
> Some architectures need special stuff (mtrr's), so I'd have to include
> .c files, too (the other way). [Look at how the table of ioctls is
> generated, its asm magic].

Shouldn't that special stuff move to the dynamic ioctl handler
registration method or the new ->compat_ioctl?

> Are you asking why are there #includes in compat_ioctl.c? Its because
> there is so many of them, and having to update all archs when you
> tuoch fs/compat_ioctl.c would be bad.

I'm asking for the #ifdef INCLUDES in fs/compat_ioctl.c.  Why do you
need it instead of including the headers uncondtionally?

Re: ioctl32: kill code duplication (sparc64 tester wanted)

[Pavel Machek]

Hi!


> > > What's the reason you can't build fs/compat_ioctl.c normally and pull
> > > in the arch magic through a magic asm/ header?  
> > 
> > Some architectures need special stuff (mtrr's), so I'd have to include
> > .c files, too (the other way). [Look at how the table of ioctls is
> > generated, its asm magic].
> 
> Shouldn't that special stuff move to the dynamic ioctl handler
> registration method or the new ->compat_ioctl?

Davem probably would not like bloat
resulting from that (There's a lot of
arch specific stuff out there). Agreed it
would be nice to do that, but I feel
we need to make small steps.


> 
> > Are you asking why are there #includes in compat_ioctl.c? Its because
> > there is so many of them, and having to update all archs when you
> > tuoch fs/compat_ioctl.c would be bad.
> 
> I'm asking for the #ifdef INCLUDES in fs/compat_ioctl.c.  Why do you
> need it instead of including the headers uncondtionally?

Because I'm including compat_ioctl.c from few places.
-- 
Horseback riding is like software...
...vgf orggre jura vgf serr.