On 11 November 2015 at 08:00, Pravin Shelar <pshe...@nicira.com> wrote:
> On Wed, Nov 11, 2015 at 12:37 AM, Joe Stringer <joestrin...@nicira.com> wrote:
>> On 10 November 2015 at 08:18, Pravin B Shelar <pshe...@nicira.com> wrote:
>>> Remove unnecessary check and definition of inet_get_local_port_range.
>>> It is already there in ip.h
>>>
>>> Reported-by: Joe Stringer <joestrin...@nicira.com>
>>> Signed-off-by: Pravin B Shelar <pshe...@nicira.com>
>>
>> Did you try this on RHEL7?
>>
> Yes, I tried it on multiple kernels - 3.10.0-229.20.1.el7,
> 3.10.0-229.14.1.el7 and 3.10.0-123.el7.

Okay, thanks for the details. I was trying on kernel
3.10.0-229.7.2.el7.x86_64. I guess that a later minor point release
changed/removed this function in the udp header.

>> Our net/ip.h defines a 3-argument version of the function. But RHEL7
>> net/udp.h uses the 2-argument version. So, we need to make sure that
>> we don't redefine it while including net/udp.h.
>>
>> AFAICT, currently any files that include net/udp.h also include
>> net/ip.h first so this logic works correctly. I suspect that the
>> lwtunnel backport introduced something which did not include net/ip.h
>> first, so the #undef logic is a no-op and the net/udp.h header would
>> fail. The patch that I sent would always include net/ip.h first, so
>> this logic would trigger and undefine our backport for the include of
>> net/udp.h, then redefine it once that was done.
>
> I do not see it with kernel that I used. This issue exist only on few
> RHEL7 releases. But it is simple fix to keep support for the kernel so
> I am fine keeping in the code. Can you send the patch you have posted
> on github?

Sure, I'll send the patch that I previously posted, and probably also
amend the comment to give a more specific detail on which RHEL version
depends on that logic.
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Reply via email to