Today ARP resolve flows (static mac-bindings) are programmed for virtual
ports once their virtual parent is claimed. As a result, during virtual
parent failover, the traffic won't switch to the new parent until the
ARP resolve flow is updated by ovn-northd, triggered by the port-binding
update.
The check_virtual_offlows_not_present() function in the case "virtual
ports" has the wrong table id 45, which should be 44. However,
correcting the table id makes the case failing, because the two ACLs
added by the case were in fact overlapping:
check ovn-nbctl acl-add sw0 to-lport 1000
Han Zhou (3):
ovn.at: Fix virtual port tests.
system-ovn.at: Add system test for virtual port with floating IP.
northd: Use dynamic mac-binding for virtual port IPs.
northd/northd.c | 103
northd/ovn-northd.8.xml | 37 ++
tests/atlocal.in
Signed-off-by: Han Zhou
---
tests/atlocal.in| 3 +
tests/system-ovn.at | 146
2 files changed, 149 insertions(+)
diff --git a/tests/atlocal.in b/tests/atlocal.in
index 0b9a312761c9..5526adac5241 100644
--- a/tests/atlocal.in
+++
On Wed, Feb 22, 2023 at 4:58 PM Lorenzo Bianconi <
lorenzo.bianc...@redhat.com> wrote:
> Lorenzo Bianconi (2):
> northd: move defrag router pipeline stage forward
> northd: add flows to defrag IP traffic
>
> northd/northd.c | 50 ++--
> northd/ovn-northd.8.xml | 60 +++--
>
I can run the fail-test more easier.
There exists a flow in verbose log like this,
recirc_id(0),in_port(2),eth(src=aa:1a:54:e9:c5:56,dst=86:29:2a:05:94:90),eth_type(0x0800),ipv4(frag=no),
packets:1, bytes:84, used:12.240s, actions:3
The env is strict. The hmap should be expanded exactly only
Bleep bloop. Greetings Aaron Conole, I am a robot and I have tried out your
patch.
Thanks for your contribution.
I encountered some error that I wasn't expecting. See the details below.
checkpatch:
WARNING: Line is 103 characters long (recommended limit is 79)
#112 FILE:
On Wed, Feb 22, 2023 at 04:19:37PM +0100, Simon Horman wrote:
> On Wed, Feb 22, 2023 at 06:33:50PM +0800, Faicker Mo wrote:
> > It's not easy to add a fail test without the changed code.
> > But I test it failed with the old code manually following these steps,
> > 1. Apply this patch(with test in
Apologies - I mis-typed Gaetan's email when I entered it into my mail
file. CC'd correctly on this email (but I can resend the patch, if you
think it is better).
Aaron Conole writes:
> Open vSwitch generally tries to let the underlying operating system
> managed the low level details of
Open vSwitch generally tries to let the underlying operating system
managed the low level details of hardware, for example DMA mapping,
bus arbitration, etc. However, when using DPDK, the underlying
operating system yields control of many of these details to userspace
for management.
In the case
Bleep bloop. Greetings Lorenzo Bianconi, I am a robot and I have tried out
your patch.
Thanks for your contribution.
I encountered some error that I wasn't expecting. See the details below.
checkpatch:
WARNING: Line is 101 characters long (recommended limit is 79)
#69 FILE:
Introduce a post_defrag router ingress pipeline stage and move current defrag
code in post_defrag stage. This is a preliminary patch to just defrag IP
fragment traffic (without performing DNAT) before accessing L4 info (e.g.
L4 protocol port) since they are not available in all IP fragment.
Introduce a priority-100 flow in the ingress router defrag stage in
order to just perform IP traffic defragmentation without doing any dnat
operation. This change is necessary since the logical flow reported
below fails for IP fragmented traffic since L4 port info is available
just in the first
Lorenzo Bianconi (2):
northd: move defrag router pipeline stage forward
northd: add flows to defrag IP traffic
northd/northd.c | 50 ++--
northd/ovn-northd.8.xml | 60 +++--
tests/ovn-northd.at | 526
tests/ovn.at| 52 ++--
Some control protocols are used to maintain link status between
forwarding engines (e.g. LACP). When the system is not sized properly,
the PMD threads may not be able to process all incoming traffic from the
configured Rx queues. When a signaling packet of such protocols is
dropped, it can cause
On Wed, Feb 22, 2023 at 06:33:50PM +0800, Faicker Mo wrote:
> It's not easy to add a fail test without the changed code.
> But I test it failed with the old code manually following these steps,
> 1. Apply this patch(with test in it)
> 2. Revert the changed code in netdev-offload-tc.c
> 3. Run the
On Wed, Feb 22, 2023 at 01:56:51PM +0100, Ales Musil wrote:
> Nothing is being freed wherever we are calling
> ctl_fatal which is fine because the program is
> about to shutdown anyway however one of the
> leaks was caught by address sanitizer.
> Fix most of the leaks that are happening before
>
On 2/20/23 19:00, Ilya Maximets wrote:
On 2/10/23 17:03, Adrián Moreno wrote:
From: Adrian Moreno
Add options to the IPFIX table configure the interval to send statistics
and template information.
Signed-off-by: Adrian Moreno
---
- v3:
- Removed unit tests which generate errors in
On Wed, Feb 22, 2023 at 07:29:51PM +0900, Nobuhiro MIKI wrote:
> When adding a route with ovs/route/add command, the source address
> in "ovs_router_entry" structure is always the FIRST address that the
> interface has. See "ovs_router_get_netdev_source_address"
> function for more information.
>
On Wed, Feb 22, 2023 at 07:29:50PM +0900, Nobuhiro MIKI wrote:
> These commands already support both IPv4 and IPv6.
>
> Signed-off-by: Nobuhiro MIKI
Reviewed-by: Simon Horman
___
dev mailing list
d...@openvswitch.org
Nothing is being freed wherever we are calling
ctl_fatal which is fine because the program is
about to shutdown anyway however one of the
leaks was caught by address sanitizer.
Fix most of the leaks that are happening before
call to ctl_fatal.
Direct leak of 64 byte(s) in 1 object(s) allocated
On Wed, Feb 22, 2023 at 1:38 PM Ilya Maximets wrote:
> +
> +if (cmd_result) {
> +free(cmd_result);
> +}
> +
> +if (cmd_error) {
> +free(cmd_error);
> +}
> >>>
> >>> I don't think the cmd_result and cmd_error conditions are
On Wed, Feb 22, 2023 at 12:30:17PM +0200, Roi Dayan via dev wrote:
> Sometimes there is a need to clean empty chains as done in
> delete_chains_from_netdev(). The cited commit doesn't remove
> the chain completely which cause adding ingress_block later to fail.
> This can be reproduced with
+
+if (cmd_result) {
+free(cmd_result);
+}
+
+if (cmd_error) {
+free(cmd_error);
+}
>>>
>>> I don't think the cmd_result and cmd_error conditions are required
>>> as free should do nothing when passed NULL.
>>>
>>
>> Isn't
On Wed, Feb 22, 2023 at 11:24:04AM +0100, Ales Musil wrote:
> On Wed, Feb 22, 2023 at 10:59 AM Simon Horman
> wrote:
...
Hi Ales,
> Hi Simon,
> thank you for your review.
you are welcome.
> > > ---
> > > utilities/ovn-dbctl.c | 76 +++
> > > 1 file
It's not easy to add a fail test without the changed code.
But I test it failed with the old code manually following these steps,
1. Apply this patch(with test in it)
2. Revert the changed code in netdev-offload-tc.c
3. Run the test
Yes, the fail-test above sometimes may pass because of the env
Sometimes there is a need to clean empty chains as done in
delete_chains_from_netdev(). The cited commit doesn't remove
the chain completely which cause adding ingress_block later to fail.
This can be reproduced with adding bond as ovs port which makes ovs
use ingress_block for it.
While at it
We can use the "ip route add ... src ..." command to set the preferred
source address for each entry in the kernel FIB. OVS has a mechanism to
cache the FIB, but the preferred source address is ignored and
calculated with its own logic. This patch resolves the difference
between kernel FIB and OVS
This patch cleans up the parser to accept pkt_mark and gw in any order.
pkt_mark and gw are normally expected to be specified exactly once.
However, as with other tools, if specified multiple times, the last
specification is used. Also, pkt_mark and gw have separate prefix
strings so they can be
When adding a route with ovs/route/add command, the source address
in "ovs_router_entry" structure is always the FIRST address that the
interface has. See "ovs_router_get_netdev_source_address"
function for more information.
If an interface has multiple ipv4 and/or ipv6 addresses, there are use
This is useful in test cases where multiple IPv4/IPv6 addresses
are assigned together.
Reviewed-by: Simon Horman
Signed-off-by: Nobuhiro MIKI
---
lib/netdev-dummy.c | 67 +-
1 file changed, 43 insertions(+), 24 deletions(-)
diff --git
With this series, the preferred source address in ovs-router is obtained
from both ovs/route/add command and kernel FIB.
v5:
- Add patch to fix man page
v4:
- Add cleanup patch for ovs/route/add
- Remove unrelated code
v3:
- Fix netdev-dummy to support multiple IP addresses
- Add validation and
These commands already support both IPv4 and IPv6.
Signed-off-by: Nobuhiro MIKI
---
ofproto/ofproto-tnl-unixctl.man | 8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/ofproto/ofproto-tnl-unixctl.man b/ofproto/ofproto-tnl-unixctl.man
index 13a465119a90..0c0a392820fb
On Wed, Feb 22, 2023 at 10:59 AM Simon Horman
wrote:
> On Wed, Feb 22, 2023 at 09:02:10AM +0100, Ales Musil wrote:
> > Nothing is being freed wherever we are calling
> > ctl_fatal which is fine because the program is
> > about to shutdown anyway however one of the
> > leaks was caught by address
On Wed, Feb 22, 2023 at 06:31:55PM +0900, Nobuhiro MIKI wrote:
> On 2023/02/22 18:04, Simon Horman wrote:
> > On Wed, Feb 22, 2023 at 05:12:36PM +0900, Nobuhiro MIKI wrote:
...
> >> diff --git a/ofproto/ofproto-tnl-unixctl.man
> >> b/ofproto/ofproto-tnl-unixctl.man
> >> index
On Wed, Feb 22, 2023 at 09:02:10AM +0100, Ales Musil wrote:
> Nothing is being freed wherever we are calling
> ctl_fatal which is fine because the program is
> about to shutdown anyway however one of the
> leaks was caught by address sanitizer.
> Fix most of the leaks that are happening before
>
On 2023/02/22 18:04, Simon Horman wrote:
> On Wed, Feb 22, 2023 at 05:12:36PM +0900, Nobuhiro MIKI wrote:
>> When adding a route with ovs/route/add command, the source address
>> in "ovs_router_entry" structure is always the FIRST address that the
>> interface has. See
On 2023/02/22 18:01, Simon Horman wrote:
> On Wed, Feb 22, 2023 at 05:12:35PM +0900, Nobuhiro MIKI wrote:
>> This patch cleans up the parser to accept pkt_mark and gw in any order.
>>
>> pkt_mark and gw are normally expected to be specified exactly once.
>> However, as with other tools, if
On Wed, Feb 22, 2023 at 10:03:07AM +0800, Faicker Mo wrote:
> Sorry.
> The commit message and code are not changed.
> Resended when I met a bug of intel-ovs-compilation test fail and add version
> descriptions.
Thanks, I think I understand now.
But please don't top-post on this mailing list.
On Wed, Feb 22, 2023 at 05:12:36PM +0900, Nobuhiro MIKI wrote:
> When adding a route with ovs/route/add command, the source address
> in "ovs_router_entry" structure is always the FIRST address that the
> interface has. See "ovs_router_get_netdev_source_address"
> function for more information.
>
On Wed, Feb 22, 2023 at 05:12:35PM +0900, Nobuhiro MIKI wrote:
> This patch cleans up the parser to accept pkt_mark and gw in any order.
>
> pkt_mark and gw are normally expected to be specified exactly once.
> However, as with other tools, if specified multiple times, the last
> specification is
This is useful in test cases where multiple IPv4/IPv6 addresses
are assigned together.
Reviewed-by: Simon Horman
Signed-off-by: Nobuhiro MIKI
---
lib/netdev-dummy.c | 67 +-
1 file changed, 43 insertions(+), 24 deletions(-)
diff --git
When adding a route with ovs/route/add command, the source address
in "ovs_router_entry" structure is always the FIRST address that the
interface has. See "ovs_router_get_netdev_source_address"
function for more information.
If an interface has multiple ipv4 and/or ipv6 addresses, there are use
We can use the "ip route add ... src ..." command to set the preferred
source address for each entry in the kernel FIB. OVS has a mechanism to
cache the FIB, but the preferred source address is ignored and
calculated with its own logic. This patch resolves the difference
between kernel FIB and OVS
With this series, the preferred source address in ovs-router is obtained
from both ovs/route/add command and kernel FIB.
v4:
- Add cleanup patch for ovs/route/add
- Remove unrelated code
v3:
- Fix netdev-dummy to support multiple IP addresses
- Add validation and unit tests for ovs/route/add
-
This patch cleans up the parser to accept pkt_mark and gw in any order.
pkt_mark and gw are normally expected to be specified exactly once.
However, as with other tools, if specified multiple times, the last
specification is used. Also, pkt_mark and gw have separate prefix
strings so they can be
Nothing is being freed wherever we are calling
ctl_fatal which is fine because the program is
about to shutdown anyway however one of the
leaks was caught by address sanitizer.
Fix most of the leaks that are happening before
call to ctl_fatal.
Direct leak of 64 byte(s) in 1 object(s) allocated
47 matches
Mail list logo