Source code compatibility in Stable series????

Source code compatibility in Stable series????

[Rogier Wolff]

Hi,

It seems that in 2.4.4 suddenly the function "skb_cow" no longer
returns the modified skb, but it retuns and integer for
succes/failure.

This means that for networking modules requiring this function, there
is no source code compatibilty between 2.4.3 and 2.4.4.

			Roger. 

-- 
** R.E.Wolff@BitWizard.nl ** http://www.BitWizard.nl/ ** +31-15-2137555 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
* There are old pilots, and there are bold pilots. 
* There are also old, bald pilots. 

Re: Source code compatibility in Stable series????

[David S. Miller]

Rogier Wolff writes:
 > It seems that in 2.4.4 suddenly the function "skb_cow" no longer
 > returns the modified skb, but it retuns and integer for
 > succes/failure.
 > 
 > This means that for networking modules requiring this function, there
 > is no source code compatibilty between 2.4.3 and 2.4.4.

And skb_datarefp went away too, in fact a ton of things changes.

Just deal with it.

Later,
David S. Miller
davem@redhat.com

Re: Source code compatibility in Stable series????

[Andi Kleen]

On Fri, May 11, 2001 at 02:56:35AM -0700, David S. Miller wrote:
> 
> Rogier Wolff writes:
>  > It seems that in 2.4.4 suddenly the function "skb_cow" no longer
>  > returns the modified skb, but it retuns and integer for
>  > succes/failure.
>  > 
>  > This means that for networking modules requiring this function, there
>  > is no source code compatibilty between 2.4.3 and 2.4.4.
> 
> And skb_datarefp went away too, in fact a ton of things changes.
> 
> Just deal with it.

I guess it would be possible to add a HAVE_ZEROCOPY to skbuff.h to make
it a bit easier for single source drivers.

--- include/linux/skbuff.h-o	Wed May  9 12:36:44 2001
+++ include/linux/skbuff.h	Fri May 11 12:12:43 2001
@@ -29,6 +29,7 @@
 #define HAVE_ALLOC_SKB		/* For the drivers to know */
 #define HAVE_ALIGNABLE_SKB	/* Ditto 8)		   */
 #define SLAB_SKB 		/* Slabified skbuffs 	   */
+#define HAVE_ZEROCOPY		/* Zerocopy stack */ 
 
 #define CHECKSUM_NONE 0
 #define CHECKSUM_HW 1


-Andi

Re: Source code compatibility in Stable series????

[Andi Kleen]

On Fri, May 11, 2001 at 12:21:59PM +0000, Petr Vandrovec wrote:
> When I was updating VMware's vmnet, I decided to use
> 
> #ifdef skb_shinfo

Yes I forgot that RedHat already shipped it :-(

> This gives you maximal backward compatibility, as all public zerocopy
> patches contain this macro. Only thing is that Dave has to remember
> that when he turns skb_shinfo into inline function, an identity #define have
> to be added.

No such guarantee for binary only software ;)

-Andi

Re: Source code compatibility in Stable series????

[Rogier Wolff]

Andi Kleen wrote:
> On Fri, May 11, 2001 at 12:21:59PM +0000, Petr Vandrovec wrote:
> > When I was updating VMware's vmnet, I decided to use
> > 
> > #ifdef skb_shinfo
> 
> Yes I forgot that RedHat already shipped it :-(
> 
> > This gives you maximal backward compatibility, as all public zerocopy
> > patches contain this macro. Only thing is that Dave has to remember
> > that when he turns skb_shinfo into inline function, an identity #define have
> > to be added.
> 
> No such guarantee for binary only software ;)

Right. I therefore refuse to maintain binary modules for my
clients. This usually convinces them to release source code.

But it's always been said that source code compatiblity would be
maintained. I'm a bit pissed that people just go about changing public
source-level interfaces. 

				Roger. 

-- 
** R.E.Wolff@BitWizard.nl ** http://www.BitWizard.nl/ ** +31-15-2137555 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
* There are old pilots, and there are bold pilots. 
* There are also old, bald pilots. 

Re: Source code compatibility in Stable series????

[Andi Kleen]

On Fri, May 11, 2001 at 12:39:29PM +0200, Rogier Wolff wrote:
> But it's always been said that source code compatiblity would be
> maintained. I'm a bit pissed that people just go about changing public
> source-level interfaces. 

2.4.4 is basically like 2.5.0 as far as networking is concerned, it includes major
fundamental changes to the stack.

-Andi

Re: Source code compatibility in Stable series????

[Rogier Wolff]

Andi Kleen wrote:
> On Fri, May 11, 2001 at 12:39:29PM +0200, Rogier Wolff wrote:
> > But it's always been said that source code compatiblity would be
> > maintained. I'm a bit pissed that people just go about changing public
> > source-level interfaces. 
> 
> 2.4.4 is basically like 2.5.0 as far as networking is concerned, it includes major
> fundamental changes to the stack.

yeah. I noticed. I'm supposed to be happy about this? 

I'll manage, and I'll "deal with it", but not before I've told you
that I'm not happy about it. 

			Roger. 

-- 
** R.E.Wolff@BitWizard.nl ** http://www.BitWizard.nl/ ** +31-15-2137555 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
* There are old pilots, and there are bold pilots. 
* There are also old, bald pilots. 

Re: Source code compatibility in Stable series????

[David S. Miller]

Andi Kleen writes:
 > I guess it would be possible to add a HAVE_ZEROCOPY to skbuff.h to make
 > it a bit easier for single source drivers.

Try MAX_SKB_FRAG, the drivers use that already.

Later,
David S. Miller
davem@redhat.com

Re: Source code compatibility in Stable series????

[David S. Miller]

Petr Vandrovec writes:
 > When I was updating VMware's vmnet, I decided to use
 > 
 > #ifdef skb_shinfo

No, don't use that, use MAX_SKB_FRAGS like the drivers do.
I guarentee to preserve that, whereas I reserve the right
to change the skb_shinfo implementation however I like.

Later,
David S. Miller
davem@redhat.com

Re: Source code compatibility in Stable series????

[David S. Miller]

Rogier Wolff writes:
 > But it's always been said that source code compatiblity would be
 > maintained.

"when possible", we've made no such total souce level
compat. guarentee.  And more such changes are coming, for example the
quota bugs can't be fixed without breaking source level compat. for
the filesystems.

You may think and argue otherwise, but our ability to break source
level compatibility is one of our strengths (see solaris rsh root
owned socket bug of yesteryear for one example as to why).

Later,
David S. Miller
davem@redhat.com

Re: Source code compatibility in Stable series????

[David S. Miller]

Andi Kleen writes:
 > 2.4.4 is basically like 2.5.0 as far as networking is concerned, it
 > includes major fundamental changes to the stack.

Andi, please.  Get over it.  That code is 6 months old.

Later,
David S. Miller
davem@redhat.com

Re: Source code compatibility in Stable series????

[Petr Vandrovec]

On 11 May 01 at 12:03, Andi Kleen wrote:
> On Fri, May 11, 2001 at 02:56:35AM -0700, David S. Miller wrote:
> I guess it would be possible to add a HAVE_ZEROCOPY to skbuff.h to make
> it a bit easier for single source drivers.
> 
> --- include/linux/skbuff.h-o    Wed May  9 12:36:44 2001
> +++ include/linux/skbuff.h  Fri May 11 12:12:43 2001
> @@ -29,6 +29,7 @@
>  #define HAVE_ALLOC_SKB     /* For the drivers to know */
>  #define HAVE_ALIGNABLE_SKB /* Ditto 8)        */
>  #define SLAB_SKB       /* Slabified skbuffs       */
> +#define HAVE_ZEROCOPY      /* Zerocopy stack */ 

When I was updating VMware's vmnet, I decided to use

#ifdef skb_shinfo

This gives you maximal backward compatibility, as all public zerocopy
patches contain this macro. Only thing is that Dave has to remember
that when he turns skb_shinfo into inline function, an identity #define have
to be added.

Just my opinion - as you cannot add HAVE_ZEROCOPY to all already existing
and installed kernels.
                                                Petr Vandrovec
                                                vandrove@vc.cvut.cz

Re: Source code compatibility in Stable series????

[Petr Vandrovec]

On 11 May 01 at 12:32, Andi Kleen wrote:
> On Fri, May 11, 2001 at 12:21:59PM +0000, Petr Vandrovec wrote:
> > When I was updating VMware's vmnet, I decided to use
> > 
> > #ifdef skb_shinfo
> 
> Yes I forgot that RedHat already shipped it :-(

Not only that RedHat shipped it, but thousands of people used Alan's
patches since 25th February. So you cannot check kernel version for >=2.4.4.
 
> > This gives you maximal backward compatibility, as all public zerocopy
> > patches contain this macro. Only thing is that Dave has to remember
> > that when he turns skb_shinfo into inline function, an identity #define have
> > to be added.
> 
> No such guarantee for binary only software ;)

vmnet is GPLed software, if you did not notice ;-) Fortunately change
went into Alan tree first, so I noticed long before 2.4.4 came out. Although
it took nonzero effort to persuade others that it will go into Linus's 2.4.x
even if it breaks API. They did not trust me that source compatibility 
in stable series is only a wish.
                                                        Petr Vandrovec
                                                        vandrove@vc.cvut.cz