Re: Chaotic structure of the net headers?

Re: Chaotic structure of the net headers?

[David S. Miller]

   From: Adrian Bunk <bunk@fs.tum.de>
   Date: Wed, 5 Mar 2003 23:54:41 +0100
   
   Two different #define's for the same thing doesn't sound like a good 
   idea?

Required by the ipv6 advanced sockets API I do believe.

Chaotic structure of the net headers?

[Adrian Bunk]

Hi,

if all I'm describing is completely logical and I'm only too dumb to see 
the logic please forgive me.  ;-)

In 2.5.64 there are networking headers both under include/linux/ and 
include/net/. I don't understand whether there's a deeper logic why e.g. 
the netfilter headers are under include/linux/.

There's some duplication, e.g. include/linux/in6.h contains

<--   snip  -->

/*
 *      IPV6 extension headers
 */
#define IPPROTO_HOPOPTS         0       /* IPv6 hop-by-hop options      */
#define IPPROTO_ROUTING         43      /* IPv6 routing header          */
#define IPPROTO_FRAGMENT        44      /* IPv6 fragmentation header    */
#define IPPROTO_ICMPV6          58      /* ICMPv6                       */
#define IPPROTO_NONE            59      /* IPv6 no next header          */
#define IPPROTO_DSTOPTS         60      /* IPv6 destination options     */

<--  snip  -->

and include/net/ipv6.h contains:

<--  snip  -->

/*
 *      NextHeader field of IPv6 header
 */

#define NEXTHDR_HOP             0       /* Hop-by-hop option header. */
#define NEXTHDR_TCP             6       /* TCP segment. */
#define NEXTHDR_UDP             17      /* UDP message. */
#define NEXTHDR_IPV6            41      /* IPv6 in IPv6 */
#define NEXTHDR_ROUTING         43      /* Routing header. */
#define NEXTHDR_FRAGMENT        44      /* Fragmentation/reassembly header. */
#define NEXTHDR_ESP             50      /* Encapsulating security payload. */
#define NEXTHDR_AUTH            51      /* Authentication header. */
#define NEXTHDR_ICMP            58      /* ICMP for IPv6. */
#define NEXTHDR_NONE            59      /* No next header */
#define NEXTHDR_DEST            60      /* Destination options header. */

<--  snip  -->

Two different #define's for the same thing doesn't sound like a good 
idea?

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: Chaotic structure of the net headers?

[David S. Miller]

   From: Rod Van Meter <Rod.VanMeter@nokia.com>
   Date: 05 Mar 2003 15:10:35 -0800
   
   Does it make sense to have two forms, one kernel, one user?  I haven't
   e.g. followed the desired include chain.  If we wanted to merge the
   uses, the former form and include location would probably have to be
   used.
   
   I've been looking into this.  There are a *few* things missing from the
   2292 support.  AFAICT, it's just a handful of functions/macros for
   manipulating option headers that need to be added.

Actually forget all my comments, GLIBC headers are where
the advanced socket API requirements for headers should
be applied.

And since this is only used in the kernel, there is no need
for the NEXTHDR_* if it trully just duplicates the IPPROTO_*
defines.

I'm willing to accept a cleanup patch of this nature, sure.

Re: Chaotic structure of the net headers?

[Rod Van Meter]

On Wed, 2003-03-05 at 14:54, ext Adrian Bunk wrote:

> 
> There's some duplication, e.g. include/linux/in6.h contains
> 
> /*
>  *      IPV6 extension headers
>  */
> #define IPPROTO_HOPOPTS         0       /* IPv6 hop-by-hop options      */
> #define IPPROTO_ROUTING         43      /* IPv6 routing header          */
> #define IPPROTO_FRAGMENT        44      /* IPv6 fragmentation header    */
> #define IPPROTO_ICMPV6          58      /* ICMPv6                       */
> #define IPPROTO_NONE            59      /* IPv6 no next header          */
> #define IPPROTO_DSTOPTS         60      /* IPv6 destination options     */


According to RFC2292 (Advanced Sockets):

2.1.1.  IPv6 Next Header Values

   IPv6 defines many new values for the Next Header field.  The
   following constants are defined as a result of including
   <netinet/in.h>.

   #define IPPROTO_HOPOPTS        0 /* IPv6 Hop-by-Hop options */
   #define IPPROTO_IPV6          41 /* IPv6 header */
   #define IPPROTO_ROUTING       43 /* IPv6 Routing header */
   #define IPPROTO_FRAGMENT      44 /* IPv6 fragmentation header */
   #define IPPROTO_ESP           50 /* encapsulating security payload */
   #define IPPROTO_AH            51 /* authentication header */
   #define IPPROTO_ICMPV6        58 /* ICMPv6 */
   #define IPPROTO_NONE          59 /* IPv6 no next header */
   #define IPPROTO_DSTOPTS       60 /* IPv6 Destination options */

   Berkeley-derived IPv4 implementations also define IPPROTO_IP to be 0.
   This should not be a problem since IPPROTO_IP is used only with IPv4
   sockets and IPPROTO_HOPOPTS only with IPv6 sockets.


> 
> and include/net/ipv6.h contains:
> 
> <--  snip  -->
> 
> /*
>  *      NextHeader field of IPv6 header
>  */
> 
> #define NEXTHDR_HOP             0       /* Hop-by-hop option header. */
> #define NEXTHDR_TCP             6       /* TCP segment. */
> #define NEXTHDR_UDP             17      /* UDP message. */
> #define NEXTHDR_IPV6            41      /* IPv6 in IPv6 */
> #define NEXTHDR_ROUTING         43      /* Routing header. */
> #define NEXTHDR_FRAGMENT        44      /* Fragmentation/reassembly header. */

This form doesn't appear in RFC2292, nor in 2133 (Basic Socket...)

My interpretation is that this latter form is defined for kernel use,
while the former is for user-level manipulation of raw packet fields
(the primary purpose of 2292).

Does it make sense to have two forms, one kernel, one user?  I haven't
e.g. followed the desired include chain.  If we wanted to merge the
uses, the former form and include location would probably have to be
used.

I've been looking into this.  There are a *few* things missing from the
2292 support.  AFAICT, it's just a handful of functions/macros for
manipulating option headers that need to be added.

Does anybody actually USE this stuff (the advanced sockets API, I mean,
not IPv6)?  I'm planning to add those missing bits, just for kicks, but
haven't done it yet.

			--Rod

Re: Chaotic structure of the net headers?

[Richard Guy Briggs]

On Wed, Mar 05, 2003 at 03:10:35PM -0800, Rod Van Meter wrote:
> On Wed, 2003-03-05 at 14:54, ext Adrian Bunk wrote:
> > There's some duplication, e.g. include/linux/in6.h contains
> > 
> > /*
> >  *      IPV6 extension headers
> >  */
> > #define IPPROTO_HOPOPTS         0       /* IPv6 hop-by-hop options      */
> > #define IPPROTO_ROUTING         43      /* IPv6 routing header          */
> > #define IPPROTO_FRAGMENT        44      /* IPv6 fragmentation header    */
> > #define IPPROTO_ICMPV6          58      /* ICMPv6                       */
> > #define IPPROTO_NONE            59      /* IPv6 no next header          */
> > #define IPPROTO_DSTOPTS         60      /* IPv6 destination options     */
> 
> 
> According to RFC2292 (Advanced Sockets):
> 
> 2.1.1.  IPv6 Next Header Values
> 
>    IPv6 defines many new values for the Next Header field.  The
>    following constants are defined as a result of including
>    <netinet/in.h>.
> 
>    #define IPPROTO_HOPOPTS        0 /* IPv6 Hop-by-Hop options */
>    #define IPPROTO_IPV6          41 /* IPv6 header */
>    #define IPPROTO_ROUTING       43 /* IPv6 Routing header */
>    #define IPPROTO_FRAGMENT      44 /* IPv6 fragmentation header */
>    #define IPPROTO_ESP           50 /* encapsulating security payload */
>    #define IPPROTO_AH            51 /* authentication header */
>    #define IPPROTO_ICMPV6        58 /* ICMPv6 */
>    #define IPPROTO_NONE          59 /* IPv6 no next header */
>    #define IPPROTO_DSTOPTS       60 /* IPv6 Destination options */
> 
>    Berkeley-derived IPv4 implementations also define IPPROTO_IP to be 0.
>    This should not be a problem since IPPROTO_IP is used only with IPv4
>    sockets and IPPROTO_HOPOPTS only with IPv6 sockets.

The Linux FreeS/WAN IPsec implementation has been using IPPROTO_ESP, 
IPPROTO_AH, IPPROTO_INT (61, put aside by IANA for internal use),
IPPROTO_COMP (108), IPPROTO_IPIP (4) for the last 5 years, based on
common usage and examples such as IPPROTO_UDP, IPPROTO_TCP, IPPROTO_ICMP.

> > and include/net/ipv6.h contains:
> > 
> > <--  snip  -->
> > 
> > /*
> >  *      NextHeader field of IPv6 header
> >  */
> > 
> > #define NEXTHDR_HOP             0       /* Hop-by-hop option header. */
> > #define NEXTHDR_TCP             6       /* TCP segment. */
> > #define NEXTHDR_UDP             17      /* UDP message. */
> > #define NEXTHDR_IPV6            41      /* IPv6 in IPv6 */
> > #define NEXTHDR_ROUTING         43      /* Routing header. */
> > #define NEXTHDR_FRAGMENT        44      /* Fragmentation/reassembly header. */
> 
> This form doesn't appear in RFC2292, nor in 2133 (Basic Socket...)
> 
> My interpretation is that this latter form is defined for kernel use,
> while the former is for user-level manipulation of raw packet fields
> (the primary purpose of 2292).

We use these in the kernel, but not in userspace.  We define SA_ESP,
etc...

> Does it make sense to have two forms, one kernel, one user?  I haven't
> e.g. followed the desired include chain.  If we wanted to merge the
> uses, the former form and include location would probably have to be
> used.

We use the two forms since shared user/kernel headers are a nuisance...

> I've been looking into this.  There are a *few* things missing from the
> 2292 support.  AFAICT, it's just a handful of functions/macros for
> manipulating option headers that need to be added.
> 
> Does anybody actually USE this stuff (the advanced sockets API, I mean,
> not IPv6)?  I'm planning to add those missing bits, just for kicks, but
> haven't done it yet.
> 
> 			--Rod

	slainte mhath, RGB

-- 
Richard Guy Briggs           --    ~\                 Auto-Free Ottawa! Canada
<www.TriColour.net>            --    \@       @           <www.flora.org/afo/>
No Internet Wiretapping!        --   _\\/\%___\\/\%        Vote! -- <Green.ca>
<www.FreeSWAN.org>_______GTVS6#790__(*)_______(*)(*)_______<www.Marillion.com>