On Wed, Jul 6, 2022 at 8:45 AM Dumitru Ceara wrote:
>
> Hi Han,
>
> On 7/6/22 00:41, Han Zhou wrote:
> > The ls_in_pre_stateful priority 120 flow that saves dst IP and Port to
> > registers is causing a critical dataplane performance impact to
> > short-lived connections, because it unwildcards
On 7/6/2022 9:57 AM, Greg Rose wrote:
It is time to remove support for the OVS kernel driver and push
towards use of the upstream Linux openvswitch kernel driver
in it's place [1].
There are many Linux specific source modules in the datapath that
will need eventual removal but some headers
On Fri, Jul 1, 2022 at 5:58 AM Mike Pattrick wrote:
>
> From: Flavio Leitner
>
> Now that there is a segmentation in software as a fall back in
> case a netdev doesn't support TCP segmentation offloading (TSO),
> enable it by default on all possible netdevs.
>
> The existing TSO control is
I did not finish reviewing.
On Fri, Jul 1, 2022 at 5:58 AM Mike Pattrick wrote:
>
> From: Flavio Leitner
>
> The netdev receiving packets is supposed to provide the flags
> indicating if the L4 csum was verified and it is OK or BAD,
> otherwise the stack will check when appropriate by software.
On Wed, Jul 6, 2022 at 5:34 PM Mike Pattrick wrote:
> > > Having to touch the packet data this late is scary, but I don't know a
> > > better place for now.
> >
> > We can move it to dp_netdev_upcall which only affects the userspace
> > datapath.
> > Then the upcall handling code gets "complete"
Thanks, Dumitru.
I pushed this to main.
On 7/6/22 09:58, Dumitru Ceara wrote:
In large scale scenarios this option hugely reduces the size of the
Southbound database positively affecting end to end performance. In
such scenarios there's no real reason to ever disable datapath groups.
In
Remove kernel based github workflows since the OVS kernel driver is
no longer supported since Release 2.18
Signed-off-by: Greg Rose
---
.github/workflows/build-and-test.yml | 53
1 file changed, 53 deletions(-)
diff --git a/.github/workflows/build-and-test.yml
The OVS kernel module is no longer supported as of OVS 2.18
Signed-off-by: Greg Rose
---
tests/automake.mk | 6 --
1 file changed, 6 deletions(-)
diff --git a/tests/automake.mk b/tests/automake.mk
index b29cb783e..3496f9002 100644
--- a/tests/automake.mk
+++ b/tests/automake.mk
@@ -343,12
As of Open vSwitch release 2.18 the OVS kernel module is no longer
supported. Pull the documentation references.
Signed-off-by: Greg Rose
---
Documentation/faq/releases.rst| 5 +-
.../contributing/backporting-patches.rst | 7 +++
Documentation/intro/install/fedora.rst
Since the openvswitch project inception it has had support for building
a Linux kernel module to support the OVS kernel datapath. Since Linux
kernel release 5.8 support for newer kernels has been deprecated. Now
is the time to fully discontinue support for building the openvswitch
kernel driver.
Remove the RHEL 6 kernel driver module specification.
Signed-off-by: Greg Rose
---
rhel/automake.mk| 5 --
rhel/kmod-openvswitch-rhel6.spec.in | 123
2 files changed, 128 deletions(-)
delete mode 100644 rhel/kmod-openvswitch-rhel6.spec.in
Remove the kernel driver specification for RHEL 7.x, 8.x and Fedora.
Signed-off-by: Greg Rose
---
rhel/automake.mk | 12 ---
rhel/openvswitch-kmod-fedora.spec.in | 152 ---
2 files changed, 164 deletions(-)
delete mode 100644
It is time to remove support for the OVS kernel driver and push
towards use of the upstream Linux openvswitch kernel driver
in it's place [1].
There are many Linux specific source modules in the datapath that
will need eventual removal but some headers are still required for
the userspace code
Hi Han,
On 7/6/22 00:41, Han Zhou wrote:
> The ls_in_pre_stateful priority 120 flow that saves dst IP and Port to
> registers is causing a critical dataplane performance impact to
> short-lived connections, because it unwildcards megaflows with exact
> match on dst IP and L4 ports. Any new
On Wed, Jul 6, 2022 at 9:02 AM David Marchand wrote:
>
> On Mon, Jul 4, 2022 at 10:24 PM David Marchand
> wrote:
> > > diff --git a/ofproto/ofproto-dpif-upcall.c b/ofproto/ofproto-dpif-upcall.c
> > > index 57f94df54..164738503 100644
> > > --- a/ofproto/ofproto-dpif-upcall.c
> > > +++
Michael Santana writes:
> On 7/4/22 09:46, Ilya Maximets wrote:
>> Hi, Michael. Thanks for the new version!
> Hi Ilya, thanks for the response
>> On 6/6/22 20:59, Michael Santana wrote:
>>> Additional threads are required to service upcalls when we have CPU
>>> isolation (in per-cpu dispatch
Timothy Redaelli writes:
> Since on CentOS/RHEL the builds are based on stable branches and not on
> tags for debugging purpose it's better to have the downstream version as
> version so it's easier to know which commits are included in a build.
>
> This commit changes the version of OVS on
On 2022-07-06 3:34 PM, Roi Dayan wrote:
On 2022-07-06 1:26 PM, Eelco Chaudron wrote:
On 5 Jul 2022, at 0:45, Ilya Maximets wrote:
TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK is supported by TC, and OVS may
create a masked match on this field.
This change is important as we're not clearing
On 2022-07-05 1:45 AM, Ilya Maximets wrote:
Current offloading code supports only limited number of tunnel keys
and silently ignores everything it doesn't understand. This is
causing, for example, offloaded ERSPAN tunnels to not work, because
flow is offloaded, but ERSPAN options are not
In large scale scenarios this option hugely reduces the size of the
Southbound database positively affecting end to end performance. In
such scenarios there's no real reason to ever disable datapath groups.
In lower scale scenarios any potential overhead due to logical datapath
groups is, very
On Mon, Jul 4, 2022 at 10:24 PM David Marchand
wrote:
> > diff --git a/ofproto/ofproto-dpif-upcall.c b/ofproto/ofproto-dpif-upcall.c
> > index 57f94df54..164738503 100644
> > --- a/ofproto/ofproto-dpif-upcall.c
> > +++ b/ofproto/ofproto-dpif-upcall.c
> > @@ -215,7 +215,7 @@ struct upcall {
> >
On 7/5/22 20:28, Gregory Rose wrote:
>
>
> On 6/29/2022 3:42 PM, Ilya Maximets wrote:
>> On 6/10/22 02:31, Frode Nordahl wrote:
>>> On Fri, Jun 3, 2022 at 4:16 PM Frode Nordahl
>>> wrote:
On Thu, Jun 2, 2022 at 6:58 PM Ilya Maximets wrote:
>
> On 6/1/22 22:53, Gregory Rose
On 2022-07-06 1:26 PM, Eelco Chaudron wrote:
On 5 Jul 2022, at 0:45, Ilya Maximets wrote:
TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK is supported by TC, and OVS may
create a masked match on this field.
This change is important as we're not clearing the masks which wasn't
really used, so if OVS
Hi
In "ovs-vsctl" utility, we have a parameter --format=json which outputs the
data in JSON format with proper semantics. I would like to add similar
"JSON" format option for 'ovs-appctl coverage/show' command. This helps the
user to consume the output data easily. We do have 2 options for this
On 5 Jul 2022, at 0:45, Ilya Maximets wrote:
> Current offloading code supports only limited number of tunnel keys
> and silently ignores everything it doesn't understand. This is
> causing, for example, offloaded ERSPAN tunnels to not work, because
> flow is offloaded, but ERSPAN options are
On 5 Jul 2022, at 0:45, Ilya Maximets wrote:
> TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK is supported by TC, and OVS may
> create a masked match on this field.
>
> This change is important as we're not clearing the masks which wasn't
> really used, so if OVS requests match on ports, we should use
On 5 Jul 2022, at 0:48, Ilya Maximets wrote:
> On 7/4/22 18:52, Eelco Chaudron wrote:
>> Hi Abhiram,
>>
>> It’s been there since your patch was committed to the repo.
>> I did a quick verification again, and it is:
>>
>> git checkout 7539b4e45
>>
>> git diff
>> diff --git
On 2022-07-06 9:53 AM, Vlad Buslov wrote:
Referenced commit changed policer action type from TC_ACT_UNSPEC (continue)
to TC_ACT_PIPE. However, since neither TC hardware offload layer nor mlx5
driver at the time validated action type and always assumed 'continue', the
breakage wasn't caught
Referenced commit changed policer action type from TC_ACT_UNSPEC (continue)
to TC_ACT_PIPE. However, since neither TC hardware offload layer nor mlx5
driver at the time validated action type and always assumed 'continue', the
breakage wasn't caught until later validation code was added. The change
29 matches
Mail list logo