>-Original Message-
>From: Ilya Maximets
>Sent: Monday, 10 June 2024 15:53
>To: Roi Dayan ; d...@openvswitch.org
>Cc: Eli Britstein ; Maor Dickman ;
>i.maxim...@ovn.org
>Subject: Re: [ovs-dev] [PATCH] dpif-netdev: Disable XPS (Transmit Packet
>Steering) for non-pmd ports.
>
>External
When using a userspace vport ("vxlan0"), dpif-netdev adds an additional
netdev ("vxlan_sys_4789"). The dpif netdev ("vxlan0") is added to the
netdev-offload ports map, thus flows are associated on this netdev.
However, flushing is done on the dpif-netdev level ("vxlan_sys_4789"),
and relevant
Vport's offloads are done on the tracked orig-in-port, but the flow itself
is associated in the vport's map.
Removing the physdev will flush all the ports that are on its map, but
not the ones on other netdevs' maps. Since flows take reference count on
both their vport and their physdev, the
When using a userspace vport ("vxlan0"), dpif-netdev adds an additional
netdev ("vxlan_sys_4789"). The dpif netdev ("vxlan0") is added to the
netdev-offload ports map, thus flows are associated on this netdev.
However, flushing is done on the dpif-netdev level ("vxlan_sys_4789"),
and relevant
Vport's offloads are done on the tracked orig-in-port, but the flow itself
is associated in the vport's map.
Removing the physdev will flush all the ports that are on its map, but
not the ones on other netdevs' maps. Since flows take reference count on
both their vport and their physdev, the
ail: Use caution opening links or attachments
>
>
>On Fri, May 26, 2023 at 4:35 PM David Marchand
> wrote:
>>
>> Hello Eli,
>>
>> On Mon, Sep 5, 2022 at 4:46 PM Eli Britstein via dev
>> wrote:
>> >
>> > Vport's offloads are d
>-Original Message-
>From: Ivan Malov
>Sent: Sunday, 4 June 2023 15:58
>To: Eli Britstein
>Cc: ovs-dev@openvswitch.org; Ilya Maximets ; Ori
>Kam ; David Marchand
>Subject: RE: [PATCH v4 3/3] netdev-offload-dpdk: use flow transfer proxy
>mechanism
>
>External email: Use caution
Hi
The kernel provides infrastructure that does not exist for DPDK. TC is
different and cannot be imitated directly in all cases.
Using a meter object created on one port (proxy) on another one (for flow) is
illegal.
Also, with this approach, even if fixed, offload will not work on if the port
Hi
The usage of the proxy port is wrong.
All transfer offload should be migrated to use it. There was [1-4], but it was
not completed.
Meters should be created only on the proxy ports. As all offloads are moved to
use proxy ports, this is how they are shared.
Since we don't know the proxy port
>-Original Message-
>From: Ivan Malov
>Sent: Tuesday, 21 February 2023 2:41
>To: ovs-dev@openvswitch.org
>Cc: Ilya Maximets ; Eli Britstein ; Ori
>Kam ; David Marchand
>Subject: [PATCH v4 3/3] netdev-offload-dpdk: use flow transfer proxy
>mechanism
>
>External email: Use caution
>-Original Message-
>From: David Marchand
>Sent: Wednesday, 11 January 2023 10:53
>To: Ivan Malov ; Eli Britstein ;
>d...@openvswitch.org
>Cc: Ilya Maximets ; Ori Kam ;
>Andrew Rybchenko ; Ian Stokes
>
>Subject: [PATCH] netdev-offload-dpdk: Fix transfer flows.
>
>External email: Use
>-Original Message-
>From: David Marchand
>Sent: Wednesday, 11 January 2023 10:32
>To: Eli Britstein ; Ivan Malov
>Cc: Ilya Maximets ; d...@openvswitch.org; Stephen
>Hemminger ; Ori Kam ;
>Maxime Coquelin ; Ian Stokes
>; Andrew Rybchenko
>
>Subject: Re: [PATCH v3 0/3] Rework the usage
>-Original Message-
>From: David Marchand
>Sent: Wednesday, 11 January 2023 9:59
>To: Ivan Malov
>Cc: Ilya Maximets ; Eli Britstein ;
>d...@openvswitch.org; Stephen Hemminger
>; Ori Kam ; Maxime
>Coquelin ; Ian Stokes
>; Andrew Rybchenko
>
>Subject: Re: [PATCH v3 0/3] Rework the usage
>-Original Message-
>From: Nole Zhang
>Sent: Monday, 9 January 2023 11:23
>To: Eli Britstein ; d...@openvswitch.org
>Cc: Eelco Chaudron ; Ilya Maximets
>; Chaoyong He ; oss-
>drivers
>Subject: RE: [PATCH 0/6] Add support for DPDK meter HW offload
>
>External email: Use caution opening
>-Original Message-
>From: Nole Zhang
>Sent: Friday, 6 January 2023 11:28
>To: Eli Britstein ; d...@openvswitch.org
>Cc: Eelco Chaudron ; Ilya Maximets
>; Chaoyong He ; oss-
>drivers ; Nole Zhang
>Subject: RE: [PATCH 0/6] Add support for DPDK meter HW offload
>
>External email: Use
Dpif-netdev should not implement internal HW offload details. If need to "apply
on all ports", it needs to be done in offload layer.
However, in arch level, there is a problem with the proposed series. It will
create a meter object per port, while in SW it is one object, that can be
shared
../ofproto/ofproto.c: In function 'handle_openflow':
../lib/ofp-bundle.c:195:15: error: 'reply.type' may be used uninitialized in
this function [-Werror=maybe-uninitialized]
195 | m->type = htons(msg->type);
| ^
../ofproto/ofproto.c:8460:36: note: 'reply.type' was
With --enable-Werror and --with-dpdk=no, with gcc (Ubuntu 11.2.0-19ubuntu1),
there are the following warnings (errors) emmitted. Those are reported
in [1] to be GCC bug. Workaround it.
In function 'memset',
inlined from 'nl_msg_put_uninit' at ../lib/netlink.c:212:9,
inlined from
With --enable-Werror and --with-dpdk=no:
../lib/dpctl.c: In function 'dpctl_ct_set_limits':
../lib/ct-dpif.c:698:22: error: 'zone' may be used uninitialized in this
function [-Werror=maybe-uninitialized]
698 | zone_limit->zone = zone;
| ^
../lib/dpctl.c:2139:18:
With --enable-Werror and --with-dpdk=no:
../ovsdb/ovsdb.c: In function 'ovsdb_schema_from_json':
../ovsdb/ovsdb.c:246:9: error: 'table' may be used uninitialized in this
function [-Werror=maybe-uninitialized]
246 | shash_add(>tables, table->name, table);
| ^
After resolving DPDK cast align warnings as stated in [1], and
resolving some more warnings in OVS side, enforce -Werror for debian and
rhel builds too.
[1] 0b6d2faace76 ("ci: Remove -Wno-cast-align from CI.")
Signed-off-by: Eli Britstein
---
debian/rules | 4 ++--
With --enable-Werror and --with-dpdk=no:
../utilities/ovs-ofctl.c: In function 'ofctl_group_mod_file':
../utilities/ovs-ofctl.c:3107:16: error: 'usable_protocols' may be used
uninitialized in this function [-Werror=maybe-uninitialized]
3107 | protocol = open_vconn_for_flow_mod(remote, ,
Fixing various warnings and enable Werror for debian/rhel builds.
Eli Britstein (8):
dp-packet: Fix dp-packet may be used initialized
ofp-port: Fix 'strnlen' specified bound may exceed source size
ovs-ofctl: Fix 'usable_protocols' may be used uninitialized
dpctl: Fix zone/limit may be
With --enable-Werror and --with-dpdk=no:
In function 'ovs_strlcpy',
inlined from 'ovs_strlcpy' at ../lib/util.c:377:1,
inlined from 'ofputil_port_to_string' at ../lib/ofp-port.c:273:9,
inlined from 'ofputil_port_from_string.part.0' at ../lib/ofp-port.c:170:13:
../lib/util.c:380:22:
With --enable-Werror and --with-dpdk=no:
In function 'dp_packet_reset_offload',
inlined from 'dp_packet_init__' at ../lib/dp-packet.c:35:5,
inlined from 'dp_packet_use__' at ../lib/dp-packet.c:50:5,
inlined from 'dp_packet_use' at ../lib/dp-packet.c:60:5,
inlined from
>-Original Message-
>From: Simon Horman
>Sent: Monday, 14 November 2022 12:44
>To: Eli Britstein
>Cc: d...@openvswitch.org; Ilya Maximets ; Simon
>Horman ; Salem Sol
>Subject: Re: [PATCH] debian, rhel: Enable Werror option in spec files
>
>External email: Use caution opening links or
Following resolving DPDK cast align warnings as stated in [1], enforce
-Werror for RPM builds too.
[1] 0b6d2faace76 ("ci: Remove -Wno-cast-align from CI.")
Signed-off-by: Eli Britstein
---
debian/rules | 4 ++--
rhel/openvswitch.spec.in | 2 +-
2 files changed, 3 insertions(+), 3
Acked-by: Eli Britstein
>-Original Message-
>From: Simon Horman
>Sent: Wednesday, 9 November 2022 21:55
>To: d...@openvswitch.org
>Cc: Eelco Chaudron ; Ilya Maximets
>; Eli Britstein ; Chaoyong He
>; oss-driv...@corigine.com; Louis Peens
>; Simon Horman
>Subject: [PATCH v4]
This series has 2 commits for DSCP and one bug fix not related to it. Not sure
why they are on the same series.
For DSCP, there is no way from RTE API point of view to rewrite partial bits on
the DSCP byte. I think those are NACK. Please explain if I'm wrong.
For the bug fix commit, the code is
When using a userspace vport ("vxlan0"), dpif-netdev adds an additional
netdev ("vxlan_sys_4789"). The dpif netdev ("vxlan0") is added to the
netdev-offload ports map, thus flows are associated on this netdev.
However, flushing is done on the dpif-netdev level ("vxlan_sys_4789"),
and relevant
Vport's offloads are done on the tracked orig-in-port, but the flow itself
is associated in the vport's map.
Removing the physdev will flush all the ports that are on its map, but
not the ones on other netdevs' maps. Since flows take reference count on
both their vport and their physdev, the
Commit [1] introduced a flag in dpif-netdev level, to optimize
performance and avoid hw_miss_packet_recover() for devices with no such
support.
However, there is a race condition between traffic processing and
assigning a 'flow_api' object to the netdev. In such case, EOPNOTSUPP is
returned by
On 8/16/2022 3:50 PM, Simon Horman wrote:
From: Chaoyong He
Add the support of offload of set IPv4 DSCP action.
Signed-off-by: Chaoyong He
Signed-off-by: Simon Horman
---
lib/netdev-offload-dpdk.c | 21 ++---
1 file changed, 18 insertions(+), 3 deletions(-)
diff --git
On 8/16/2022 3:50 PM, Simon Horman wrote:
From: Chaoyong He
Add support of IPv6 tunnel.
IPv6 tunnel is already supported. The missing setting of is_ipv6 field
affects only if using a dpdk PMD that would return restore info flags
with RTE_FLOW_RESTORE_INFO_TUNNEL on, but
>-Original Message-
>From: Ivan Malov
>Sent: Wednesday, July 20, 2022 3:18 PM
>To: d...@openvswitch.org
>Cc: Eli Britstein ; Stephen Hemminger
>; Ilya Maximets ; Ori
>Kam ; Maxime Coquelin
>; David Marchand
>; Andrew Rybchenko
>
>Subject: [PATCH v3 3/3] netdev-offload-dpdk: use flow
A connection is established if we see packets from both directions.
The cited commit [1] fixed the issue of sending twice in one direction,
but still an issue if more than that.
Fix it.
Fixes: a867c010ee91 ("conntrack: Fix conntrack new state")
Signed-off-by: Eli Britstein
---
Commit [1] introduced a flag in dpif-netdev level, to optimize
performance and avoid hw_miss_packet_recover() for devices with no such
support.
However, there is a race condition between traffic processing and
assigning a 'flow_api' object to the netdev. In such case, EOPNOTSUPP is
returned by
+Ori
From: Finn, Emma
Sent: Friday, June 24, 2022 5:41 PM
To: Ilya Maximets ; ovs-dev@openvswitch.org
; Stokes, Ian
Cc: Eli Britstein ; Slava Ovsiienko ;
Flavio Leitner ; Matan Azrad
Subject: RE: [ovs-dev] [PATCH] netdev-offload-dpdk: setting RSS hash types
Hi Ivan,
>-Original Message-
>From: Ivan Malov
>Sent: Wednesday, June 8, 2022 10:02 PM
>To: Eli Britstein
>Cc: d...@openvswitch.org; Andrew Rybchenko
>; Ilya Maximets ;
>Ori Kam ; NBU-Contact-Thomas Monjalon (EXTERNAL)
>; Stephen Hemminger
>; David Marchand
>; Gaetan Rivet ; Maxime
Hi Ivan,
>-Original Message-
>From: Ivan Malov
>Sent: Wednesday, June 8, 2022 5:46 PM
>To: Eli Britstein
>Cc: d...@openvswitch.org; Andrew Rybchenko
>; Ilya Maximets ;
>Ori Kam ; NBU-Contact-Thomas Monjalon (EXTERNAL)
>; Stephen Hemminger
>; David Marchand
>; Gaetan Rivet ; Maxime
Hi Ivan,
>-Original Message-
>From: Ivan Malov
>Sent: Tuesday, June 7, 2022 11:56 PM
>To: Eli Britstein
>Cc: d...@openvswitch.org; Andrew Rybchenko
>; Ilya Maximets ;
>Ori Kam ; NBU-Contact-Thomas Monjalon (EXTERNAL)
>; Stephen Hemminger
>; David Marchand
>; Gaetan Rivet ; Maxime
"TUNNEL_ID" is a bad name, but that's how dpdk called it.
There was a discussion about having this knowledge in OVS so we can avoid
calling rte_flow_get_restore_info(). How else it is used?
>-Original Message-
>From: Ivan Malov
>Sent: Monday, May 30, 2022 5:16 PM
>To:
- Missing proper handling of the testpmd syntax logging. It changes the used
port according to "transfer", but the log still uses netdev_dpdk_get_port_id().
- The usage of the "proxy" port for rte-flow implies that this proxy port is
attached to OVS, otherwise it is not "started" and creation of
gating the offload design?
2. I have not seen companion patch support for encap offload for Geneve. Is
similar restriction of not offloading Geneve w/ options apply on the encap
offload?
Hemal
On Sat, May 7, 2022 at 10:07 PM Eli Britstein via dev
mailto:ovs-dev@openvswitch.org>>
>-Original Message-
>From: Ilya Maximets
>Sent: Wednesday, May 4, 2022 2:44 PM
>To: Eli Britstein ; d...@openvswitch.org
>Cc: i.maxim...@ovn.org; Gaetan Rivet ;
>msant...@redhat.com; Nir Anteby
>Subject: Re: [PATCH V2 3/3] netdev-offload-dpdk: Add geneve header pattern
>match
>
Tunnel offload APIs have '__rte_experimental' attribute, therefore
available only if ALLOW_EXPERIMENTAL_API is defined. Documente it.
Signed-off-by: Eli Britstein
---
Documentation/howto/dpdk.rst | 4
1 file changed, 4 insertions(+)
diff --git a/Documentation/howto/dpdk.rst
attribute
>
> External email: Use caution opening links or attachments
>
>
> On 2/7/22 17:56, Eli Britstein via dev wrote:
> > DPDK 20.11 introduced an ability to specify existance/non-existance of
> > VLAN tag by [1].
> > Use this attribute.
> >
> >
Add support for matching on geneve header.
Signed-off-by: Eli Britstein
Reviewed-by: Nir Anteby
Acked-by: Michael Santana
---
NEWS | 2 ++
lib/netdev-offload-dpdk.c | 58 +++
2 files changed, 60 insertions(+)
diff --git a/NEWS b/NEWS
Add support for dpdk geneve tunnel offloads.
v2-v1:
- Rebase.
GitHub actions:
v1: https://github.com/elibritstein/OVS/actions/runs/1633157455
v2: https://github.com/elibritstein/OVS/actions/runs/1807027490
Eli Britstein (3):
netdev-dpdk: Add flow_api support for netdev geneve vports
Signed-off-by: Eli Britstein
Reviewed-by: Nir Anteby
Acked-by: Michael Santana
---
lib/netdev-offload-dpdk.c | 13 +
1 file changed, 13 insertions(+)
diff --git a/lib/netdev-offload-dpdk.c b/lib/netdev-offload-dpdk.c
index 94dc6a9b7..edd4e009d 100644
---
Add the acceptance of geneve devices to netdev_dpdk_flow_api_supported()
API, to allow offloading of DPDK geneve devices.
Signed-off-by: Eli Britstein
Reviewed-by: Nir Anteby
Acked-by: Michael Santana
---
lib/netdev-dpdk.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git
A flow may be modified after its initial offload failed. In this case,
according to [1], the modification is handled as a flow add.
For a vport flow "add", the orig_in_port should be provided.
Keep that field in the flow struct, so it can be provided in the flow
modification use case.
[1]
For VLANs, the match of ethernet type should be specified in inner_type
field of the vlan match, and not type field in ethernet match.
Fix it.
Fixes: e8a2b5bf92bb ("netdev-dpdk: implement flow offload with rte flow")
Signed-off-by: Eli Britstein
Reviewed-by: Salem Sol
---
DPDK 20.11 introduced an ability to specify existance/non-existance of
VLAN tag by [1].
Use this attribute.
[1]: 09315fc83861 ("ethdev: add VLAN attributes to ethernet and VLAN items")
Signed-off-by: Eli Britstein
Reviewed-by: Salem Sol
---
lib/netdev-offload-dpdk.c | 16
1
On 1/4/2022 4:56 PM, Eelco Chaudron wrote:
External email: Use caution opening links or attachments
On 5 Dec 2021, at 8:34, Eli Britstein via dev wrote:
Following [1]-[3] in DPDK, there are no more such warnings from DPDK.
Remove ignoring them if they occur.
GitHub actions:
v1: https
Add support for matching on geneve header.
Signed-off-by: Eli Britstein
Reviewed-by: Nir Anteby
---
NEWS | 2 ++
lib/netdev-offload-dpdk.c | 58 +++
2 files changed, 60 insertions(+)
diff --git a/NEWS b/NEWS
index bc4a1cfac..18f15b61f
Signed-off-by: Eli Britstein
Reviewed-by: Nir Anteby
---
lib/netdev-offload-dpdk.c | 13 +
1 file changed, 13 insertions(+)
diff --git a/lib/netdev-offload-dpdk.c b/lib/netdev-offload-dpdk.c
index 402353125..50650bd07 100644
--- a/lib/netdev-offload-dpdk.c
+++
Add support for dpdk geneve tunnel offloads.
GitHub actions:
v1: https://github.com/elibritstein/OVS/actions/runs/1633157455
Eli Britstein (3):
netdev-dpdk: Add flow_api support for netdev geneve vports
netdev-offload-dpdk: Support tnl_pop for geneve tunnel
netdev-offload-dpdk: Add geneve
Add the acceptance of geneve devices to netdev_dpdk_flow_api_supported()
API, to allow offloading of DPDK geneve devices.
Signed-off-by: Eli Britstein
Reviewed-by: Nir Anteby
---
lib/netdev-dpdk.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/lib/netdev-dpdk.c
On 12/17/2021 12:06 AM, Ilya Maximets wrote:
External email: Use caution opening links or attachments
New 'options:dummy-rte-flow' configuration knob for netdev-dpdk
ports, i.e. type=dpdk/dpdkvhostuserclient, to turn on dummy rte_flow
API. This version of API will do nothing but reply with
Following [1]-[3] in DPDK, there are no more such warnings from DPDK.
Remove ignoring them if they occur.
GitHub actions:
v1: https://github.com/elibritstein/OVS/actions/runs/1540651133
[1] a3f8d0587188 ("net: avoid cast-align warning in VLAN insert function")
[2] da0333c8790b ("mbuf: avoid
Acked-by: Eli Britstein
On 12/3/2021 11:12 PM, Ilya Maximets wrote:
External email: Use caution opening links or attachments
Last RX queue, from which the packet got received, is already stored
in the PMD context. So, we can get the netdev from it without the
expensive hash map lookup.
In
On 11/30/2021 2:06 PM, Ilya Maximets wrote:
External email: Use caution opening links or attachments
On 11/30/21 10:43, Eli Britstein wrote:
On 11/30/2021 12:31 AM, Ilya Maximets wrote:
External email: Use caution opening links or attachments
On 11/7/21 11:56, Eli Britstein via dev wrote
On 11/30/2021 12:31 AM, Ilya Maximets wrote:
External email: Use caution opening links or attachments
On 11/7/21 11:56, Eli Britstein via dev wrote:
Following dpdk commits [1]-[3], it is now possible to compile with
--enable-Werror. Change the default to on, with an option to disable
using
On 11/22/2021 3:19 PM, Sriharsha Basavapatna wrote:
Hi Eli,
On Sun, Nov 21, 2021 at 12:03 PM Eli Britstein via dev
wrote:
Hi Harsha,
It's a clever idea, though have some problems in the implementation. PSB.
Thanks, please see my response below.
On 11/20/2021 11:20 AM, Sriharsha
Hi Harsha,
It's a clever idea, though have some problems in the implementation. PSB.
On 11/20/2021 11:20 AM, Sriharsha Basavapatna wrote:
The hw_miss_packet_recover() API results in performance degradation, for
ports that are either not offload capable or do not support this specific
offload
Following dpdk commits [1]-[3], it is now possible to compile with
--enable-Werror. Change the default to on, with an option to disable
using --disable-Werror.
Notes:
1. To compile against 21.11-rc1, need to apply [4] and [5] patches.
2. There are still sparse errors, due to dpdk issue. [6] fixes
On 8/27/2021 12:30 PM, Maxime Coquelin wrote:
External email: Use caution opening links or attachments
On 8/16/21 3:53 PM, Eli Britstein via dev wrote:
Signed-off-by: Eli Britstein
---
NEWS | 2 ++
lib/netdev-offload-dpdk.c | 5 -
2 files changed, 2 insertions
On 8/24/2021 6:47 PM, Ilya Maximets wrote:
External email: Use caution opening links or attachments
On 8/24/21 5:25 PM, Eli Britstein wrote:
On 8/24/2021 6:08 PM, Finn, Emma wrote:
External email: Use caution opening links or attachments
-Original Message-
From: Eli Britstein
On 8/24/2021 6:08 PM, Finn, Emma wrote:
External email: Use caution opening links or attachments
-Original Message-
From: Eli Britstein
Sent: Monday 16 August 2021 14:55
To: d...@openvswitch.org; Ilya Maximets
Cc: Finn, Emma ; Stokes, Ian ; Sriharsha Basavapatna
; Gaetan Rivet ;
A flow may be modified after its initial offload failed. In this case,
according to [1], the modification is handled as a flow add.
For a vport flow "add", the orig_in_port should be provided.
Keep that field in the flow struct, so it can be provided in the flow
modification use case.
[1]
For VLANs, the match of ethernet type should be specified in inner_type
field of the vlan match, and not type field in ethernet match.
Fix it.
Fixes: e8a2b5bf92bb ("netdev-dpdk: implement flow offload with rte flow")
Signed-off-by: Eli Britstein
Reviewed-by: Salem Sol
---
DPDK 20.11 introduced an ability to specify existance/non-existance of
VLAN tag by [1].
Use this attribute.
[1]: 09315fc83861 ("ethdev: add VLAN attributes to ethernet and VLAN items")
Signed-off-by: Eli Britstein
Reviewed-by: Salem Sol
---
lib/netdev-offload-dpdk.c | 16
1
Support IPv6 fragmentation matching.
Signed-off-by: Eli Britstein
---
lib/netdev-offload-dpdk.c | 82 ++-
1 file changed, 81 insertions(+), 1 deletion(-)
diff --git a/lib/netdev-offload-dpdk.c b/lib/netdev-offload-dpdk.c
index ee8e56fe0..d79ad1bea 100644
---
Signed-off-by: Eli Britstein
---
NEWS | 2 ++
lib/netdev-offload-dpdk.c | 5 -
2 files changed, 2 insertions(+), 5 deletions(-)
diff --git a/NEWS b/NEWS
index 26920e215..80466c014 100644
--- a/NEWS
+++ b/NEWS
@@ -8,6 +8,8 @@ Post-v2.16.0
by default.
Matching on frag types requires range. Add 'last' attribute to patterns.
Signed-off-by: Eli Britstein
---
lib/netdev-offload-dpdk.c | 151 --
1 file changed, 81 insertions(+), 70 deletions(-)
diff --git a/lib/netdev-offload-dpdk.c b/lib/netdev-offload-dpdk.c
Support IPv4 fragmentation matching.
Signed-off-by: Eli Britstein
---
lib/netdev-offload-dpdk.c | 47 +--
1 file changed, 45 insertions(+), 2 deletions(-)
diff --git a/lib/netdev-offload-dpdk.c b/lib/netdev-offload-dpdk.c
index 769dc48d2..ee8e56fe0 100644
Netdev datapath ignored fragmentation matching so far. [1] introduced
an API for that in dpdk. Use it to apply frag matching.
[1] http://mails.dpdk.org/archives/dev/2020-October/186177.html
Travis:
v1: https://travis-ci.org/github/elibritstein/OVS/builds/752634689
GitHub Actions:
v1:
On 8/2/2021 10:58 PM, Ilya Maximets wrote:
External email: Use caution opening links or attachments
On 7/26/21 10:14 AM, Eli Britstein wrote:
This patch-set improves/fixes offloads transitions behavior.
Patch #1 avoids flushing PMD offloads unnecessarily.
Patch #2 fixes a flow
Association of a mark to a flow is done as part of its offload handling,
in the offloading thread. However, the PMD thread specifies whether an
offload request is an "add" or "modify" by the association of a mark to
the flow.
This is exposed to a race condition. A flow might be created with
Log flow modifications to help debugging.
Signed-off-by: Eli Britstein
Reviewed-by: Gaetan Rivet
---
lib/dpif-netdev.c | 114 +++---
1 file changed, 68 insertions(+), 46 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index
dp_netdev_flow_offload_main thread is asynchronous, by the cited commit.
There might be a case where there are modification requests of the same
flow submitted before handled. Then, if the first handling fails, the
rule for the flow is deleted, and the mark is freed. Then, the following
one should
Before flushing offloads of a removed port was supported by [1], it was
necessary to flush the 'marks'. In doing so, all offloads of the PMD are
removed, include the ones that are not related to the removed port and
that are not modified following this removal. As a result such flows are
evicted
This patch-set improves/fixes offloads transitions behavior.
Patch #1 avoids flushing PMD offloads unnecessarily.
Patch #2 fixes a flow modifications bug.
Patch #3 fixes a race condition with flow modifications.
Patch #3 improves debuggability of flow modifications.
v2-v1:
- Rebase.
v3-v2:
-
Reported-by: Harry Van Haaren
Fixes: 4e432d6f8128 ("netdev-offload-dpdk: Support tnl/push using vxlan encap
attribute.")
Fixes: e098c2f966cb ("netdev-dpdk-offload: Add vxlan pattern matching
function.")
Signed-off-by: Eli Britstein
---
lib/netdev-offload-dpdk.c | 19 +--
1
Fixes: b6207b1d2711 ("netdev-offload-dpdk: Support offload of set IPv6
actions.")
Signed-off-by: Eli Britstein
---
lib/netdev-offload-dpdk.c | 5 -
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/lib/netdev-offload-dpdk.c b/lib/netdev-offload-dpdk.c
index e7913292e..4112fc3a5
On 7/23/2021 9:00 PM, Ilya Maximets wrote:
External email: Use caution opening links or attachments
On 7/12/21 5:07 PM, Eli Britstein wrote:
Before flushing offloads of a removed port was supported by [1], it was
necessary to flush the 'marks'. In doing so, all offloads of the PMD are
On 7/22/2021 4:10 PM, Ilya Maximets wrote:
External email: Use caution opening links or attachments
On 7/22/21 3:00 PM, Eli Britstein wrote:
On 7/22/2021 3:28 PM, Ilya Maximets wrote:
External email: Use caution opening links or attachments
On 7/11/21 7:15 AM, Eli Britstein wrote:
On 7/22/2021 3:28 PM, Ilya Maximets wrote:
External email: Use caution opening links or attachments
On 7/11/21 7:15 AM, Eli Britstein wrote:
Compiling with -Werror and -Wcast-align has errors like:
lib/netdev-offload-dpdk.c: In function 'dump_flow_pattern':
lib/netdev-offload-dpdk.c:385:38:
On 7/15/2021 4:35 PM, Ferriter, Cian wrote:
External email: Use caution opening links or attachments
-Original Message-
From: Eli Britstein
Sent: Wednesday 14 July 2021 16:21
To: Ferriter, Cian ; Ilya Maximets
; Gaëtan Rivet
; d...@openvswitch.org; Van Haaren, Harry
Cc: Majd
90 matches
Mail list logo
