Re: MARK bit operations patch #2

[Anders Fugmann <afu@fugmann.dhs.org>]

Harald Welte wrote:
> On Tue, Jan 14, 2003 at 01:52:17PM +0100, Anders Fugmann wrote:
> 
>>The patch is compatible in the following ways:
>>
>>new kernel - old userspace: OK
>>new kernel - old userspace: OK
> 
> 
> twice the same? 
My bad, should have said:
new kernel - new userspace: OK
old kernel - new userspace: OK
new kernel - old userspace: OK(*)
* - Only if the new userspace is compiled against the old kernel.

(And of course: old kernel - old userspace :-))

>>
>>Please comment. I'm not use if the compatibility satifies the 
>>requirements needed for such a patch to go into the 2.4 series.
> 
> 
> I think this is fine.  The major cases we have to think about are:
> 
> 1) user updates kernel (new kernel, old userspace)
>    The firewall should start up as expected without any changes in
>    userspace
Works.
> 
> 2) user has old kenel running and is updating iptables userspace
>    In this case iptables userspace has to compile without any warnings.
Also Works.
> 
> 
>>Attached: MARK_TARGET.patch - patch against 2.4 (applies to 2.5 aswell))
>>Attached: MARK_operations.patch.userspace  - patch against iptables .1.2.7a)
> 
> 
> Looks fine so far.  The only issue remaining is:
> 
> You are assuming that 
> 
> IPT_ALIGN(sizeof(struct ipt_mark_target_info)) 
> 	!= IPT_ALIGN(sizeof(struct ipt_mark_bitops_target_info))
> 
> while this may be true on x86, I assume it is different on 64bit
> plaforms. Do you have access to any ia64/sparc64/ppc64/mips64 machine
> with linux?
Unfortunatly not. But why could this introduce an problem on a non x86
platform? Also do you know where to look for more information on this?

I also forgot the manpage update. It will come in the next batch.
Thanks for the positive feedback.

Regards
Anders Fugmann