On 05/14/2014 05:55 PM, Andrey Korolyov wrote:
On 05/14/2014 12:56 AM, Jesse Gross wrote:
On Tue, May 13, 2014 at 1:52 PM, Andrey Korolyov <and...@xdel.ru> wrote:
On Wed, May 14, 2014 at 12:44 AM, Jesse Gross <je...@nicira.com> wrote:
On Tue, May 13, 2014 at 1:25 PM, Andrey Korolyov <and...@xdel.ru> wrote:
On Wed, May 14, 2014 at 12:19 AM, Jesse Gross <je...@nicira.com> wrote:
On Tue, May 13, 2014 at 1:58 AM, Andrey Korolyov <and...@xdel.ru> wrote:
On Tue, May 13, 2014 at 4:41 AM, Jesse Gross <je...@nicira.com> wrote:
On Mon, May 12, 2014 at 2:04 PM, Andrey Korolyov <and...@xdel.ru> wrote:
On Tue, May 13, 2014 at 12:58 AM, Jesse Gross <je...@nicira.com> wrote:
On Sat, May 10, 2014 at 2:13 PM, Andrey Korolyov <and...@xdel.ru> wrote:
Hello,
sorry for bringing this old thread on top, but right now building
against current RHEL 3.10.121.el7 seems not possible without
patchwork. After fixing some obvious kversion checks and code
snippets, I had stuck on f067987973facf28916a2f5ee48bf02ce5938bf8:
In file included from
/builder/rpmbuild/BUILD/openvswitch-2.1.3/_default/../datapath/linux/compat/include/net/gre.h:9:0,
from
/builder/rpmbuild/BUILD/openvswitch-2.1.3/_default/datapath/linux/vport-gre.c:44:
include/net/gre.h: In function ‘gre_handle_offloads’:
include/net/gre.h:42:2: error: implicit declaration of function
‘iptunnel_handle_offloads’ [-Werror=implicit-function-declaration]
return iptunnel_handle_offloads(skb, gre_csum, SKB_GSO_GRE);
^
include/net/gre.h:42:2: warning: return makes pointer from integer
without a cast [enabled by default]
cc1: some warnings being treated as errors
Bare look says that this snippet should be modified more intensively
than by one-line patch to work well, so I thought it would be better
to ask there on how to fix this properly.
What version of OVS is this and which patch are you specifically referring to?
Sorry if it was unclear, I pointed hash above, it` s 2.1.2. By patch I
meant one-liners which helps to mock kernel version checking in the
places where conflict of backported functions can take place. I
replaced failed checks one by one and then stuck with error above.
Actually, both gre.h - from compat/ and from RH kernel headers
resulting to this error if one want to try to point them explicitly.
Can you post the changes you had to make to get to this point? I would
expect that HAVE_GRE_CISCO_REGISTER would be defined and so there
should be relatively little that is version specific.
Have you also tried branch-2.2 or master?
There are no such thing in public repo as branch-2.2, so with master
75264fe5f1ceb73a33bcc8e1adb3ad9aad57237d I`ve got the same. Please
take a look on the attached patch. Also mind that the change in
datapath.c is purely a mock to pass through it to next failure and
shouldn`t go anywhere :)
OK - I see. These changes are just miscellaneous backports and not in
the tunnel code as I thought you were saying. I don't have a copy of
the RHEL 7 beta source handy but it seems like this could be a missing
definition in those headers. OVS should just be using the tunnel code
from the kernel headers and the use of iptunnel_handle_offloads()
would be from there. If they use it, they should also define it...
You can use srpm for headers:
http://ftp.redhat.com/redhat/rhel/rc/7/ComputeNode/source/tree/Packages/kernel-3.10.0-121.el7.src.rpm
Since you're already working in this code, it might be faster for you
to take a look.
Ok, will check tomorrow morning, but since RH-Oracle counterstand took
place, there are no reliable changelog except text remarks in the spec
file, so it`s a bit uneasy to say which bits should be presented and
which should not...
I would just look into how and where iptunnel_handle_offloads is used
and defined. I don't think that there is really more to it than that.
OVS is using the GRE receive code wholesale in this version so there
shouldn't be any tricky dependencies.
I had modified checks a little, so right now module compiles but there
are unported fixes after 3.11 for ip_tunnel stuff which are doing build
warnings. Also differences in declaration of skb_zerocopy (void vs
non-void) making proper build impossible without porting all code
related to this function from RH`s version of queue_userspace_packet to
mainline using a lot of ``if'' blocks. Please take a look on a revised
patch and attached build log. I`ve added some folks in CC who are known
for work on OVS subsystem of RedHat kernel, so they may have a word on
which changes will be incorporated in next minor kernel release and
resulting patch can be created without wasting much effort away.
Unfortunately Zoltan's fix affecting skb_zerocopy() was merged after el7
code freeze so 7.0 will export that as void but you'll need a workaround
anyway as upstream kernels have been released for both signatures.
A subsequent update is very likely to incorporate the fix which will
change the return value to int.
_______________________________________________
discuss mailing list
discuss@openvswitch.org
http://openvswitch.org/mailman/listinfo/discuss
