Re: [ovs-dev] [PATCH v3] dpif-netdev: fix vlan and ipv4 parsing in avx512

Hi Harry, Tested Again and looks ohk now in random testing. Regards Amber > -Original Message- > From: Van Haaren, Harry > Sent: Monday, January 31, 2022 7:25 PM > To: ovs-dev@openvswitch.org > Cc: Ferriter, Cian ; Stokes, Ian > ; i.maxim...@ovn.org; echau...@redhat.com; Amber, > Kumar

Re: [ovs-dev] [PATCH ovn] northd: Add feature to log reply and related ACL traffic.

Bleep bloop. Greetings Mark Michelson, 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 80 characters long (recommended limit is 79) #62 FILE:

[ovs-dev] [PATCH ovn] northd: Add feature to log reply and related ACL traffic.

It can be desirable for replies to stateful ACLs to be logged. And in some cases, it can actually be a bit confusing why they aren't logged. Consider a situation where a port group called "port_group" exists and logical switch ports swp1 and swp2 belong to it. We create the following ACL, where

[ovs-dev] [PATCH 1/1] odp-util: Fix tunnel key attr for GTP-U.

CC: William Tu Fixes: 3c6d05a02e0f ("userspace: Add GTP-U support.") Signed-off-by: Nobuhiro MIKI --- lib/odp-util.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/lib/odp-util.c b/lib/odp-util.c index 503675e83..9a705cffa 100644 --- a/lib/odp-util.c +++ b/lib/odp-util.c @@

Re: [ovs-dev] [PATCH] ovsdb-idl: ovsdb_idl_loop_destroy must also destroy the committing txn.

On 1/20/22 14:04, Dumitru Ceara wrote: > Found by AddressSanitizer when running OVN tests: > Direct leak of 64 byte(s) in 1 object(s) allocated from: > #0 0x498fb2 in calloc > (/home/runner/work/ovn/ovn/ovn-21.12.90/_build/sub/ic/ovn-ic+0x498fb2) > #1 0x5f681e in xcalloc__

Re: [ovs-dev] [PATCH v2 08/18] python: add ovs datapath flow parsing

Adrian Moreno writes: > diff --git a/python/ovs/flows/odp.py b/python/ovs/flows/odp.py > new file mode 100644 > index 0..a86a0d6b1 > --- /dev/null > +++ b/python/ovs/flows/odp.py > @@ -0,0 +1,783 @@ [...] > +class ODPFlow(Flow): [...] > +def __init__(self, odp_string, id=None): >

Re: [ovs-dev] [PATCH] conntrack: Check TCP state while testing established connections pick up.

On 1/27/22 00:44, Gaëtan Rivet wrote: > On Mon, Dec 27, 2021, at 14:26, Paolo Valerio wrote: >> When testing if an established connection is picked up, it could be >> useful to verify that the protocol state matches the expectation, that >> is, it moves to ESTABLISHED, as there's a chance that

Re: [ovs-dev] [PATCH v2 0/2] Improve userspace Tx steering documenation.

On 1/24/22 17:58, Maxime Coquelin wrote: > This series is a follow-up of "Add missing tx-steering PMD > option." patch. The second patch fixes incorrect statement on > static/dynamic thread-to-txq mapping in thread mode. > > Changes in v2: > == > - Add Documentation patch to fix

Re: [ovs-dev] [PATCH v3 0/2] Fixes for mpls encap/decap actions

On 1/20/22 05:05, Martin Varghese wrote: > From: Martin Varghese > > This series contains fixes for the commit > 1917ace89364 ("Encap & Decap actions for MPLS packet type."). > > Martin Varghese (2): > tests: Fix cosmetic errors in system-traffic.at. > ofproto-dpif-xlate: Fix packet drops

Re: [ovs-dev] [PATCH 0/3] Fix IDL condition processing when reconnecting.

On 1/11/22 17:37, Dumitru Ceara wrote: > Dumitru Ceara (3): > python: idl: Resend requested but not acked conditions when > reconnecting. > ovsdb-cs: Clear last_id on reconnect if condition changes in-flight. > python: idl: Clear last_id on reconnect if condition changes

Re: [ovs-dev] [PATCH] tests: Add de-serialization check to the json string benchmark.

On 1/12/22 14:04, Aaron Conole wrote: > Ilya Maximets writes: > >> Since we're testing serialization, it also makes sense to test >> the opposite operation. Should be useful in the future for >> exploring possible optimizations. >> >> CMD: $ ./tests/ovstest json-string-benchmark >> >>

Re: [ovs-dev] [PATCH v4] ofp-flow: skip flow reply if it exceeds the maximum message size

On 1/10/22 09:13, Eelco Chaudron wrote: > Currently, if a flow reply results in a message which exceeds > the maximum reply size, it will assert OVS. This would happen > when OVN uses OpenFlow15 to add large flows, and they get read > using OpenFlow10 with ovs-ofctl. > > This patch prevents this

Re: [ovs-dev] [PATCH 2/4] ovsdb: transaction: Keep one entry in the transaction history.

On 1/25/22 02:34, Han Zhou wrote: > > > On Sun, Dec 19, 2021 at 6:09 AM Ilya Maximets > wrote: >> >> If a single transaction exceeds the size of the whole database (e.g., >> a lot of rows got removed and new ones added), transaction history will >> be drained.  This

Re: [ovs-dev] [PATCH 1/4] ovsdb-cs: Fix ignoring of the last id from the initial monitor reply.

On 1/25/22 02:27, Han Zhou wrote: > > > On Sun, Dec 19, 2021 at 6:09 AM Ilya Maximets > wrote: >> >> Current code doesn't use the last id received in the monitor reply. >> That may result in re-downloading the database content if the >> re-connection happened after

Re: [ovs-dev] [PATCH RFC 4/5] conntrack: Split single cmap to multiple buckets.

Hi Gaetan, thank you for the feedback Gaëtan Rivet writes: > Hi Paolo, > > On Mon, Nov 29, 2021, at 19:06, Paolo Valerio wrote: >> The purpose of this commit is to split the current way of storing the >> conn nodes. Before this patch the nodes were stored into a single cmap >> using ct->lock

Re: [ovs-dev] [PATCH v2 07/18] python: introduce OpenFlow Flow parsing

Adrian Moreno writes: > diff --git a/python/ovs/flows/ofp.py b/python/ovs/flows/ofp.py > new file mode 100644 > index 0..1af06fa01 > --- /dev/null > +++ b/python/ovs/flows/ofp.py > @@ -0,0 +1,453 @@ [...] > + > +class OFPFlow(Flow): [...] > +def __init__(self, ofp_string,

Re: [ovs-dev] [PATCH v2 06/18] python: add flow base class

Adrian Moreno writes: > diff --git a/python/ovs/flows/flow.py b/python/ovs/flows/flow.py > new file mode 100644 > index 0..2456d5f87 > --- /dev/null > +++ b/python/ovs/flows/flow.py > @@ -0,0 +1,125 @@ [...] > +class Flow(object): > +"""The Flow class is a base class for other

Re: [ovs-dev] [PATCH v2 03/18] python: add list parser

Adrian Moreno writes: > diff --git a/python/ovs/flows/list.py b/python/ovs/flows/list.py > new file mode 100644 > index 0..cafd23d0a > --- /dev/null > +++ b/python/ovs/flows/list.py > @@ -0,0 +1,124 @@ > +import re > + > +from ovs.flows.kv import KeyValue, KeyMetadata, ParseError > +from

Re: [ovs-dev] [PATCH] dpif-netdev: Fix a race condition in deletion of offloaded flows

On 1/31/22 11:42, Gaëtan Rivet wrote: > On Thu, Jan 27, 2022, at 11:38, Ilya Maximets wrote: >> On 1/27/22 07:42, Sriharsha Basavapatna via dev wrote: >>> In dp_netdev_pmd_remove_flow() we schedule the deletion of an >>> offloaded flow, if a mark has been assigned to the flow. But if >>> this

Re: [ovs-dev] [PATCH v2 01/18] python: add generic Key-Value parser

Adrian Moreno writes: > diff --git a/python/ovs/flows/decoders.py b/python/ovs/flows/decoders.py > new file mode 100644 > index 0..73d5c0c60 > --- /dev/null > +++ b/python/ovs/flows/decoders.py > @@ -0,0 +1,18 @@ > +""" Defines helpful decoders that can be used to decode information from

Re: [ovs-dev] [PATCH v2 02/18] python: add mask, ip and eth decoders

James Troup writes: >> +Both base10 and base16 integers are supported. > > Is this true? If `value` is a string, it isn't. int("0x4") throws an > exception but int(0x4) doesn't. Ah, sorry, I see now. int("0x34") => ValueError, int("0x34, 0) => 52. -- James

Re: [ovs-dev] [PATCH v2 02/18] python: add mask, ip and eth decoders

Adrian Moreno writes: > diff --git a/python/ovs/flows/decoders.py b/python/ovs/flows/decoders.py > index 73d5c0c60..2f8e5bd0a 100644 > --- a/python/ovs/flows/decoders.py > +++ b/python/ovs/flows/decoders.py > @@ -16,3 +25,392 @@ def decode_default(value): > return int(value, 0) >

Re: [ovs-dev] [PATCH ovn] vtep: set is-vtep to chassis's other_config if absent

Hi Dumitru, thanks for the review. I’ve addressed requested changes in v2: https://patchwork.ozlabs.org/project/ovn/patch/20220131173457.64391-1-odiv...@gmail.com/ Regards, Vladislav Odintsov > On 28 Jan 2022, at 20:06, Dumitru Ceara wrote: > > On 1/28/22 18:04, Dumitru Ceara wrote: >> On

[ovs-dev] [PATCH ovn v2] vtep: set is-vtep to chassis's other_config if absent

After commit [0] the mandatory vtep chassis's option 'is-vtep' has appeared. The upgrade scenario for ovn-controller-vtep was not supported: 'is-vtep' option was set only on vtep chassis creation. If chassis was created prior to a new codebase, it was left intact and HW VTEP connectivity was

Re: [ovs-dev] [PATCH v2] tc: Fix incorrect TC rule for decap+encap datapath flow

On 31 Jan 2022, at 14:58, Roi Dayan wrote: > On 2021-12-21 10:55 AM, Roi Dayan wrote: >> >> >> On 2021-12-12 9:45 AM, Roi Dayan wrote: >>> >>> >>> On 2021-12-02 2:38 PM, Roi Dayan wrote: A datapath flow generated for traffic from vxlan port to another vxlan port looks like this:

Re: [ovs-dev] [PATCH v2] tc: Fix incorrect TC rule for decap+encap datapath flow

On 2021-12-21 10:55 AM, Roi Dayan wrote: On 2021-12-12 9:45 AM, Roi Dayan wrote: On 2021-12-02 2:38 PM, Roi Dayan wrote: A datapath flow generated for traffic from vxlan port to another vxlan port looks like this:

[ovs-dev] [PATCH v3] dpif-netdev: fix vlan and ipv4 parsing in avx512

This commit fixes the minimum packet size for the vlan/ipv4/tcp traffic profile, which was previously incorrectly set. This commit also disallows any fragmented IPv4 packets from being matched in the optimized miniflow-extract, avoiding complexity of handling fragmented packets and using scalar

Re: [ovs-dev] [PATCH v2 10/10] odp-util: Fix output for tc to be equal to kernel

On 2022-01-31 12:55 PM, Eelco Chaudron wrote: When the same flow is programmed in the kernel and tc, they look different due to the way they are translated. They take the userspace approach by always including the packet type attribute. To make the outputs the same, show the ethernet header

Re: [ovs-dev] [PATCH v2 09/10] revalidator: Fix datapath statistics update.

On 2022-01-31 12:54 PM, Eelco Chaudron wrote: Make sure to only update packet and byte counters when valid, or else this could lead to "temporarily/occasionally" out-of-sync flow counters. push_dp_ops() will now handle updating the stats similar to the way it's handled in revalidate_ukey().

Re: [ovs-dev] [PATCH v2 08/10] netdev-offload-tc: Check for none offloadable ct_state flag combination

On 2022-01-31 12:53 PM, Eelco Chaudron wrote: This patch checks for none offloadable ct_state match flag combinations. If they exist tell the user about it once, and ignore the offload. Signed-off-by: Eelco Chaudron --- lib/netdev-offload-tc.c | 18 +++--- 1 file changed,

Re: [ovs-dev] [PATCH v2 07/10] netdev-offload-tc: Fix IP and port ranges in flower returns.

On 2022-01-31 12:52 PM, Eelco Chaudron wrote: When programming NAT rules OVS only sets the minimum value for a single IP/port value. However, responses from flower will always return min == max for single IP/port values. This is causing the verification to fail as the request is different

Re: [ovs-dev] [PATCH v2 06/10] netdev-offload-tc: Fix use of ICMP values instead of masks defines.

On 2022-01-31 12:38 PM, Eelco Chaudron wrote: Signed-off-by: Eelco Chaudron --- lib/tc.c | 13 + 1 file changed, 5 insertions(+), 8 deletions(-) diff --git a/lib/tc.c b/lib/tc.c index 8cfd5aa5a..ebec097dc 100644 --- a/lib/tc.c +++ b/lib/tc.c @@ -939,24 +939,21 @@

Re: [ovs-dev] [PATCH v2 05/10] netdev-offload-tc: Always include conntrack information to tc

On 2022-01-31 12:37 PM, Eelco Chaudron wrote: Regardless of the traffic type, if requested, the conntrack information should be included to keep the datapath and tc rules in sync. Signed-off-by: Eelco Chaudron --- lib/tc.c | 10 +- 1 file changed, 5 insertions(+), 5

Re: [ovs-dev] [PATCH v2 04/10] netdev-offload-tc: Check for valid netdev ifindex in flow_put

On 2022-01-31 12:34 PM, Eelco Chaudron wrote: Verify that the returned ifindex by netdev_get_ifindex() is valid. This might not be the case in the ERSPAN port scenario, which can not be offloaded. Signed-off-by: Eelco Chaudron --- lib/netdev-offload-tc.c |9 + 1 file changed,

Re: [ovs-dev] [PATCH v2 03/10] odp_util: Fix parse_key_and_mask_to_match() vlan parsing

On 2022-01-31 12:34 PM, Eelco Chaudron wrote: The parse_key_and_mask_to_match() is a function to translate a netlink formatted key/mask to match structure. And should not consider any configuration setting when translating. Signed-off-by: Eelco Chaudron --- lib/odp-util.c | 41

Re: [ovs-dev] [PATCH v2 02/10] netdev-offload-tc: Set the correct VLAN_VID and VLAN_PCP masks

On 2022-01-31 12:33 PM, Eelco Chaudron wrote: This change will set the correct VID and PCP masks, as well as the ethernet type mask. Signed-off-by: Eelco Chaudron --- v2: Fixed sparse "expected restricted ovs_be16" warning lib/tc.c | 10 ++ 1 file changed, 6 insertions(+), 4

Re: [ovs-dev] [PATCH v2 01/10] netdev-offload-tc: Add debug logs on tc rule verify failures

On 2022-01-31 12:32 PM, Eelco Chaudron wrote: This patch adds more detailed debug logs on tc verify failures to ease debugging the actual cause after the fact. Signed-off-by: Eelco Chaudron --- lib/tc.c | 79 +- 1 file

Re: [ovs-dev] [PATCH v2] dpif-netdev: fix vlan and ipv4 parsing in avx512

> -Original Message- > From: Eelco Chaudron > Sent: Monday, January 31, 2022 10:05 AM > To: Van Haaren, Harry > Cc: ovs-dev@openvswitch.org; Ferriter, Cian ; Stokes, > Ian ; i.maxim...@ovn.org > Subject: Re: [PATCH v2] dpif-netdev: fix vlan and ipv4 parsing in avx512 > > Only one real

[ovs-dev] [PATCH v1] system-dpdk.at: Improve mfex test to include dpif avx512.

AVX512 DPIF must be active in order for the MFEX AutoValidator to be executed. If the DPIF-AVX512 is not available, the unit test is skipped, as the scalar DPIF does not use the MFEX function-pointer based optimizations. Suggested-by: Cian Ferriter Signed-off-by: Kumar Amber ---

Re: [ovs-dev] [PATCH 09/10] netdev-offload-tc: stats should be captured on first action in the list

On 2022-01-31 12:16 PM, Eelco Chaudron wrote: On 30 Jan 2022, at 14:30, Roi Dayan wrote: On 2022-01-27 3:35 PM, Eelco Chaudron wrote: Statistics should be captured on the first action in a list of actions, to do this we need to also capture the last update time stamp from the first

[ovs-dev] [PATCH v2 10/10] odp-util: Fix output for tc to be equal to kernel

When the same flow is programmed in the kernel and tc, they look different due to the way they are translated. They take the userspace approach by always including the packet type attribute. To make the outputs the same, show the ethernet header when the packet type is wildcarded, and not printed.

[ovs-dev] [PATCH v2 09/10] revalidator: Fix datapath statistics update.

Make sure to only update packet and byte counters when valid, or else this could lead to "temporarily/occasionally" out-of-sync flow counters. push_dp_ops() will now handle updating the stats similar to the way it's handled in revalidate_ukey(). Signed-off-by: Eelco Chaudron ---

[ovs-dev] [PATCH v2 08/10] netdev-offload-tc: Check for none offloadable ct_state flag combination

This patch checks for none offloadable ct_state match flag combinations. If they exist tell the user about it once, and ignore the offload. Signed-off-by: Eelco Chaudron --- lib/netdev-offload-tc.c | 18 +++--- 1 file changed, 15 insertions(+), 3 deletions(-) diff --git

[ovs-dev] [PATCH v2 07/10] netdev-offload-tc: Fix IP and port ranges in flower returns.

When programming NAT rules OVS only sets the minimum value for a single IP/port value. However, responses from flower will always return min == max for single IP/port values. This is causing the verification to fail as the request is different than the response. To avoid this, we will update the

Re: [ovs-dev] [PATCH] dpif-netdev: Fix a race condition in deletion of offloaded flows

On Thu, Jan 27, 2022, at 11:38, Ilya Maximets wrote: > On 1/27/22 07:42, Sriharsha Basavapatna via dev wrote: >> In dp_netdev_pmd_remove_flow() we schedule the deletion of an >> offloaded flow, if a mark has been assigned to the flow. But if >> this occurs in the window in which the offload thread

[ovs-dev] [PATCH v2 06/10] netdev-offload-tc: Fix use of ICMP values instead of masks defines.

Signed-off-by: Eelco Chaudron --- lib/tc.c | 13 + 1 file changed, 5 insertions(+), 8 deletions(-) diff --git a/lib/tc.c b/lib/tc.c index 8cfd5aa5a..ebec097dc 100644 --- a/lib/tc.c +++ b/lib/tc.c @@ -939,24 +939,21 @@ nl_parse_flower_ip(struct nlattr **attrs, struct tc_flower

[ovs-dev] [PATCH v2 05/10] netdev-offload-tc: Always include conntrack information to tc

Regardless of the traffic type, if requested, the conntrack information should be included to keep the datapath and tc rules in sync. Signed-off-by: Eelco Chaudron --- lib/tc.c | 10 +- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/lib/tc.c b/lib/tc.c index

[ovs-dev] [PATCH v2 04/10] netdev-offload-tc: Check for valid netdev ifindex in flow_put

Verify that the returned ifindex by netdev_get_ifindex() is valid. This might not be the case in the ERSPAN port scenario, which can not be offloaded. Signed-off-by: Eelco Chaudron --- lib/netdev-offload-tc.c |9 + 1 file changed, 9 insertions(+) diff --git

[ovs-dev] [PATCH v2 03/10] odp_util: Fix parse_key_and_mask_to_match() vlan parsing

The parse_key_and_mask_to_match() is a function to translate a netlink formatted key/mask to match structure. And should not consider any configuration setting when translating. Signed-off-by: Eelco Chaudron --- lib/odp-util.c | 41 - 1 file changed, 28

[ovs-dev] [PATCH v2 02/10] netdev-offload-tc: Set the correct VLAN_VID and VLAN_PCP masks

This change will set the correct VID and PCP masks, as well as the ethernet type mask. Signed-off-by: Eelco Chaudron --- v2: Fixed sparse "expected restricted ovs_be16" warning lib/tc.c | 10 ++ 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/lib/tc.c b/lib/tc.c index

[ovs-dev] [PATCH v2 01/10] netdev-offload-tc: Add debug logs on tc rule verify failures

This patch adds more detailed debug logs on tc verify failures to ease debugging the actual cause after the fact. Signed-off-by: Eelco Chaudron --- lib/tc.c | 79 +- 1 file changed, 73 insertions(+), 6 deletions(-) diff --git

[ovs-dev] [PATCH v2 00/10] netdev-offload-tc: Fix various tc-offload related problems

This series fixes a bunch of TC offload-related issues found when running the kernel data path self-test (make check-kernel) forcing TC to be enabled. To do this manually I applied the following diff: https://github.com/chaudron/ovs/commit/d7ff1060f371e2cbd4b2e8dee222a9eed55073c8 These

Re: [ovs-dev] [PATCH 09/10] netdev-offload-tc: stats should be captured on first action in the list

On 30 Jan 2022, at 14:30, Roi Dayan wrote: > On 2022-01-27 3:35 PM, Eelco Chaudron wrote: >> Statistics should be captured on the first action in a list >> of actions, to do this we need to also capture the last update >> time stamp from the first action. > > Why stats should be captured on the

Re: [ovs-dev] dev Digest, Vol 150, Issue 161

Hi Harry, Thanks for the fix while testing found an issue around frag masking. The mask is causing one bit to be not exactly what is being generated from scalar MFEX and due to which Auto-validator tests are failing.

Re: [ovs-dev] [PATCH v2] dpif-netdev: fix vlan and ipv4 parsing in avx512

Only one real concern about the -1 usage, the rest looks fine (just some questions). //Eelco On 28 Jan 2022, at 16:20, Harry van Haaren wrote: > This commit fixes the minimum packet size for the vlan/ipv4/tcp > traffic profile, which was previously incorrectly set. > > This commit also