netdev - Re: [PATCH] openvswitch: Make IPv6 packet parsing dependent on IPv6 config

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]

Date:	Fri, 16 Nov 2012 10:46:17 -0800
From:	Jesse Gross <jesse@...ira.com>
To:	vyasevic@...hat.com
Cc:	dev@...nvswitch.org, netdev@...r.kernel.org, davem@...emloft.net,
	fengguang.wu@...el.com
Subject: Re: [PATCH] openvswitch: Make IPv6 packet parsing dependent on IPv6 config

On Fri, Nov 16, 2012 at 9:43 AM, Vlad Yasevich <vyasevic@...hat.com> wrote:
> On 11/16/2012 12:26 PM, Jesse Gross wrote:
>>
>> On Fri, Nov 16, 2012 at 7:40 AM, Vlad Yasevich <vyasevic@...hat.com>
>> wrote:
>>
>>> Openvswitch attempts to use IPv6 packet parsing functions without
>>> any dependency on IPv6 (unlike every other place in kernel).  Pull
>>> the IPv6 code in openvswitch togeter and put a conditional that's
>>> dependent on CONFIG_IPV6.
>>>
>>> Resolves:
>>> net/built-in.o: In function `ovs_flow_extract':
>>> (.text+0xbf5d5): undefined reference to `ipv6_skip_exthdr'
>>>
>>> Signed-off-by: Vlad Yasevich <vyasevic@...hat.com>
>>
>>
>>
>> Doesn't this move in the opposite direction of your patches to make IPv6
>> GSO/GRO always available?  The packets being processed here
>> are generally created by the guest but with Open vSwitch running on the
>> host.  Also, ipv6_skip_exthdr() is in exthdrs_core.c, so it actually is
>> always available.  I suspect that the real problem is that the dependency
>> on the ipv6 directory changed to CONFIG_INET and Open vSwitch should now
>> depend on this.
>>
>
> Yes and no... :)  IPv6 uses a bunch of IPv4 code all over.  IPv4 is enabled
> with CONFIG_INET and IPv6 with CONFIG_NET.  So creates a strange imbalance.
> By shifting IPv6 to CONFIG_INET (which is where it
> lives and what enables its selection during config process), we now have a
> dependency with openvswitch.
>
> All other users of ipv6_skip_exthdr have it either under the IS_ENABLED
> conditional or through some other means that don't build it when INET is
> completely turned off.  This patch does the same for openvswitch.
>
> I see 2 alternatives to this:
>  1) Make openvswitch depend on CONFIG_INET.
>  2) Pull a ton of code out of CONFIG_INET (v4 and v6) and into CONFIG_NET.
> This could start with IPv6 header parsing and maybe even
> include GSO/TSO (but not sure how much sense that would be).
>
> What's your take?

I agree the IPv4 and IPv6 code is all tangled together and that IPv6
should use CONFIG_INET as well.  I think in an ideal world we would
separate them out but it seems like a lot of work for not much
practical benefit.

I think if you took this to the logical extension and restricted all
the protocols based on the kernel config (i.e. IPv4, TCP, and UDP are
conditional on CONFIG_INET) then you end up with a confusing mess.  On
the other hand, if you do it only for IPv6 it's also confusing because
the fact that the other protocols are simple enough to parse on their
own and IPv6 is more complicated is really an implementation detail
that shouldn't be exposed.

I guess the simplest thing to do seems to just make Open vSwitch
depend on CONFIG_INET seeing as it is practically useless without
upper layer protocol support anyways.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html