On 6 March 2017 at 16:22, Jarno Rajahalme <[email protected]> wrote:
> This patch set backports the recent upstream conntrack fixes and new
> features to the OVS tree kernel module, and adds the OVS userspace
> support.
>
> Patch 1/22 is an unrelated datapath backport, and patch 22/22 allows
> compiling against Linux 4.10.
>
> Each new feature is introduced in two different commits, the first is
> the datapath backport, the second the corresponding userspace datapath
> and non-datapath functionality, including OVS system tests.  In one
> instance I have squashed the system test with the datapath backport.
> Compile would fail after the first patch due to missing userspace code
> for new enums.  We may decide to squash the datapath and userspace
> changes together for the merge, but for now the review should be more
> straightforward with the separation.
>
> System tests have been most recently run on Linux 3.16, on which the
> geneve tests fail, but that should have nothing to do with this
> series.
>
> v3: Address Joe's feedback.

I saw some minor issues which I commented on - primarily style and one
functional question in patch #15, otherwise I see a few areas for
improvement:

* System-traffic tests do not check for ct_orig_tuple support before
running, so on kernels without support the new tests fail
* There is no OpenFlow-layer rejection of matches on ct_orig_tuple
OXMs, or resubmit(,,ct) when there is no datapath support for
ct_orig_tuple.
* Similarly ct(force) doesn't have any OpenFlow-visible error
reporting based on datapath support

These could be addressed in subsequent patches, so long as they make
it before 2.8.

For the whole series:
Acked-by: Joe Stringer <[email protected]>
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to