netdev - Re: [PATCH net] ipv6: set all.accept

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <b2df5d25-f9f9-c08e-b3f7-db8f1bb38110@6wind.com>
Date:   Mon, 13 Nov 2017 17:09:33 +0100
From:   Nicolas Dichtel <nicolas.dichtel@...nd.com>
To:     Stefano Brivio <sbrivio@...hat.com>,
        Maciej Żenczykowski <maze@...gle.com>
Cc:     Erik Kline <ek@...gle.com>, David Miller <davem@...emloft.net>,
        netdev <netdev@...r.kernel.org>, Matteo Croce <mcroce@...hat.com>
Subject: Re: [PATCH net] ipv6: set all.accept_dad to 0 by default

Le 13/11/2017 à 16:05, Stefano Brivio a écrit :
> On Mon, 13 Nov 2017 23:52:26 +0900
> Maciej Żenczykowski <maze@...gle.com> wrote:
> 
>>>> Should we consider rolling back the patch that caused this?
>>>> "accept_dad = 1" is the proper IETF-expected default behaviour.
>>>>
>>>> Alternatively, if we really want to make all, default, and ifname
>>>> useful perhaps we need to investigate a tristate option (for currently
>>>> boolean values, at least).  -1 could mean no preference, for example.  
>>>
>>> I haven't checked how ugly it would be, yet. But another way to restore
>>> the previous behaviour, while keeping the new functionality, would be
>>> to keep the global default as 1 and instead set the per-interface
>>> accept_dad default value to 0. What do you think?  
>>
>> The default out-of-the-box behaviour should definitely be to do DAD.
Yes, and my patch didn't modify this.

>>
>> You can achieve this in 4 ways:
>>
>> [A] all=1, default=1, AND --> the OLD pre-patch behaviour
> 
> Old pre-patch behaviour simply ignored the 'all' value though.
> 
>> [B] all=1, default=1, OR --> the NEW post-patch behaviour - problematic
>> [C] all=1, default=0, OR --> problematic for same reason: iface=0 is a no-op
> 
> But this way you could still globally disable DAD, starting from
> default values, by simply setting 'all' to zero, which is what Nicolas
> wanted.
> 
>> [D] all=0, default=1, OR
>>
>> Note that:
>> AND == (all < 1 || interface < 1)
>> OR == (all < 1 && interface < 1)
>>
>> [C] requires one to set all but one interface (incl. default) to 1,
>> then set all=0,
>> just to disable a single interface's dad
>>
>> [D] is weird, because with the default already being dad enabled, there's really
>> no reason to ever set all=1
>>
>> Being able to disable either for all interfaces (via all=0) or for a
>> specific interface (via iface=0) seems
>> the most useful.
>>
>> Setting all=1, default=0, specific_interfaces=1, AND-logic also seems useful.
>>
>> Hence my vote to rollback a2d3f3e33853.
> 
> We're mostly talking about 35e015e1f577 here.
> 
I don't have a strong opinion about what to do.

My reasoning was that before patch 35e015e1f577, all.accept_dad had no effect,
thus I took the assumption that users didn't modify it, but only
default.accept_dad and <iface>.accept_dad. With this assumption, my patch helps
to keep the same settings when upgrading the kernel.

Nicolas