Re: [PATCH] fix iptables on systems with discontiguous processor ids

["Sébastien Bernard" <seb@frankengul.org>]

David S. Miller a écrit :

>From: seb@frankengul.org
>Date: Wed, 12 Oct 2005 08:22:18 +0200
>
>  
>
>>After building, the modules_install reports the symbol undefined.
>>Well, I see now that it is only a lack of EXPORT_SYMBOL.
>>    
>>
>
>Aha, that makes sense.
>
>  
>
>>All smp arch used cpu_online_map and cpu_possible_map but sparc which
>>used phys_cpu_present_map (a specificity of the sparc port I presume).
>>    
>>
>
>Yes, we do this because since we lack cpu hotplug support
>this means "cpu present"=="cpu possible".
>
>  
>
Ok, heres is the oops provoked by the second patch posted by Harald on 
the 11th of October.

unable to handle paging_request (some zeros...)140098000

iptables_restor(2488): Oops [#1]
TPC: <get_couters +0x68/0xe0>

Caller: do_ipt_get_ctl + 0x5b8/0x7a0
            nf_sock_opt + 0xa4 / 0x160
            ip_get_sockopt + 0xa4 / 0x5e0
            sock_common_getsockopt + 0x1c/0x40
            sys_get_sockopt + 0x84/0xe0
            compat_sys_getsockopt + 0x28/0x120
            linux_sparc_syscall32 + 0x34/0x40

Note that it is a different crash.
The previous was a unable to handle kernel request NULL dereference.
As a result, the machine locks hard.

Hope it helps.

    Seb