/usr/src/linux/Documentation/networking/ifenslave.c

/usr/src/linux/Documentation/networking/ifenslave.c

[Mads Martin Jørgensen]

$SUBJECT is a userspace tool, but includes kernel headers. This small
obvious patch against 2.6.0-test1 should fix it.

--- ifenslave.c
+++ ifenslave.c
@@ -97,8 +97,7 @@
 #include <sys/types.h>
 #include <sys/socket.h>
 #include <sys/ioctl.h>
-#include <linux/if.h>
-#include <linux/if_arp.h>
+#include <net/if_arp.h>
 #include <linux/if_ether.h>
 #include <linux/if_bonding.h>
 #include <linux/sockios.h>

-- 
Mads Martin Joergensen, http://mmj.dk
"Why make things difficult, when it is possible to make them cryptic
 and totally illogical, with just a little bit more effort?"
                                -- A. P. J.

Re: /usr/src/linux/Documentation/networking/ifenslave.c

[David S. Miller]

On Wed, 16 Jul 2003 11:23:54 +0200
Mads Martin Jørgensen <mmj@suse.de> wrote:

> $SUBJECT is a userspace tool, but includes kernel headers. This small
> obvious patch against 2.6.0-test1 should fix it.

This patch is absolutely senseless, you eliminate two such includes

> -#include <linux/if.h>
> -#include <linux/if_arp.h>
> +#include <net/if_arp.h>

but three more still remain.

>  #include <linux/if_ether.h>
>  #include <linux/if_bonding.h>
>  #include <linux/sockios.h>

This whole anti-kernel-headers-in-userspace thing is a
total shamans dance and not founded in reality.

Quick, answer this as fast as you can, where in the glibc headers can
you get at the PFKEY and XFRM_NETLINK interface definitions to
configure IPSEC stuff in the kernel?

BZZT, time is up, and I know you have no answer :-)

That is why all of this is rediculious.

Re: /usr/src/linux/Documentation/networking/ifenslave.c

[Mads Martin Jørgensen]

* David S. Miller <davem@redhat.com> [Jul 16. 2003 14:52]:
> This whole anti-kernel-headers-in-userspace thing is a
> total shamans dance and not founded in reality.

Urgs.

> Quick, answer this as fast as you can, where in the glibc headers can
> you get at the PFKEY and XFRM_NETLINK interface definitions to
> configure IPSEC stuff in the kernel?
> 
> BZZT, time is up, and I know you have no answer :-)

:-)

> That is why all of this is rediculious.

Problem is some of these includes are different on different archs, and
causes the thing to miscompile. How to fix that then?

-- 
Mads Martin Joergensen, http://mmj.dk
"Why make things difficult, when it is possible to make them cryptic
 and totally illogical, with just a little bit more effort?"
                                -- A. P. J.

Re: /usr/src/linux/Documentation/networking/ifenslave.c

[David S. Miller]

On Wed, 16 Jul 2003 14:59:10 +0200
Mads Martin Jørgensen <mmj@suse.de> wrote:

> > That is why all of this is rediculious.
> 
> Problem is some of these includes are different on different archs, and
> causes the thing to miscompile. How to fix that then?

Nothing arch specific resides in linux/if.h :-)

This means the problem eminates from asm/*.h headers which is
where the fixes belong.

What exactly is the error you get on ia64?

Re: /usr/src/linux/Documentation/networking/ifenslave.c

[Mads Martin Jørgensen]

* David S. Miller <davem@redhat.com> [Jul 16. 2003 15:21]:
> > > That is why all of this is rediculious.
> > 
> > Problem is some of these includes are different on different archs, and
> > causes the thing to miscompile. How to fix that then?
> 
> Nothing arch specific resides in linux/if.h :-)
> 
> This means the problem eminates from asm/*.h headers which is
> where the fixes belong.

Yes. What fix, would you propose?

> What exactly is the error you get on ia64?

a bunch of compile errors like the following:

In file included from /usr/src/linux/include/asm/system.h:19,
                 from /usr/src/linux/include/asm/atomic.h:17,
                 from /usr/src/linux/include/linux/netdevice.h:32,
                 from /usr/src/linux/include/linux/if_arp.h:26,
                 from ifenslave.c:91:
/usr/src/linux/include/asm/pal.h:89: parse error before "pal_status_t"
/usr/src/linux/include/asm/pal.h:89: warning: type defaults to `int' in 
declaration of `pal_status_t'
/usr/src/linux/include/asm/pal.h:89: warning: data definition has no
type or storage class
/usr/src/linux/include/asm/pal.h:102: parse error before
"pal_cache_level_t"
/usr/src/linux/include/asm/pal.h:102: warning: type defaults to `int' in 

declaration of `pal_cache_level_t'
/usr/src/linux/include/asm/pal.h:102: warning: data definition has no
type or 
storage class
/usr/src/linux/include/asm/pal.h:110: parse error before
"pal_cache_type_t"
/usr/src/linux/include/asm/pal.h:110: warning: type defaults to `int' in 
declaration of `pal_cache_type_t'
/usr/src/linux/include/asm/pal.h:110: warning: data definition has no
type or storage class
/usr/src/linux/include/asm/pal.h:123: parse error before
"pal_cache_line_state_t

etc.

thanks,
-- 
Mads Martin Joergensen, http://mmj.dk
"Why make things difficult, when it is possible to make them cryptic
 and totally illogical, with just a little bit more effort?"
                                -- A. P. J.

Re: /usr/src/linux/Documentation/networking/ifenslave.c

[David S. Miller]

On Wed, 16 Jul 2003 15:35:14 +0200
Mads Martin Jørgensen <mmj@suse.de> wrote:

> * David S. Miller <davem@redhat.com> [Jul 16. 2003 15:21]:
> In file included from /usr/src/linux/include/asm/system.h:19,
>                  from /usr/src/linux/include/asm/atomic.h:17,
>                  from /usr/src/linux/include/linux/netdevice.h:32,
>                  from /usr/src/linux/include/linux/if_arp.h:26,
>                  from ifenslave.c:91:
> /usr/src/linux/include/asm/pal.h:89: parse error before "pal_status_t"

IA64 needs it's __KERNEL__ ifdefs fixed up, nothing more.

Thanks for finding the right place to fix this problem :-)

Re: /usr/src/linux/Documentation/networking/ifenslave.c

[Jeff Garzik]

On Wed, Jul 16, 2003 at 05:42:55AM -0700, David S. Miller wrote:
> This whole anti-kernel-headers-in-userspace thing is a
> total shamans dance and not founded in reality.

There ARE definitions that differ between kernel and userspace.

Thinking they are the same is not founded in reality.

And thinking that all of userspace is instantly recompiled against
against the latest kernel headers isn't founded in reality, either.

	Jeff

Re: /usr/src/linux/Documentation/networking/ifenslave.c

[David S. Miller]

On Wed, 16 Jul 2003 10:12:06 -0400
Jeff Garzik <jgarzik@pobox.com> wrote:

> There ARE definitions that differ between kernel and userspace.

Sure, but none of the one's were talking about here.

> Thinking they are the same is not founded in reality.

I'm not.

> And thinking that all of userspace is instantly recompiled against
> against the latest kernel headers isn't founded in reality, either.

GLIBC is munging netlink messages on the way out via sendmsg()?
That'd be news to me. :-)