On Thu, Aug 25, 2022 at 10:05 AM Han Zhou wrote:
>
>
>
> On Thu, Aug 25, 2022 at 7:47 AM Xavier Simonart
wrote:
> >
> > If a logical switch port is added and connected to a logical router
> > port (through options: router-port) before the router port is
> > created, then this might cause further
This patch adds the necessary support to avx512 mfex to
support handling of tunnel packet type.
Signed-off-by: Kumar Amber
---
v5:
- check metadata IP address to find tunneling is valid or not.
As dummy-pmd often passes garbage data to dpif.
---
---
lib/dpif-netdev-avx512.c | 16
The Mfex study function is split into outer and inner to allow
for independent selection and studying of packets in outer and inner
flows to different ISA optimized Mfexs.
Signed-off-by: Kumar Amber
---
lib/dpif-netdev-extract-study.c | 126 +---
1 file changed, 83
The md_is_valid parameter is passed from DPIF to MFEX to allow mfex
functions to detect the tunneling and decide the processing of Inner
packets in static predictable branches.
Signed-off-by: Kumar Amber
---
lib/dpif-netdev-avx512.c | 3 ++-
lib/dpif-netdev-extract-avx512.c | 9
The set command in MFEX is changed as to allow the user to select
different optimized mfex ISA for processing Inner packets in case
of tunneling.
$ ovs-appctl dpif-netdev/miniflow-parser-set -pmd 3 study 1024 -recirc
The get command is modified to indcitate both inner and Outer MFEXs in
use.
The patch adds and re-uses the dpif set command to set the
function pointers to be used to switch between different inner
dpifs.
Signed-off-by: Kumar Amber
Signed-off-by: Cian Ferriter
Co-authored-by: Cian Ferriter
---
v3:
- Add description to the dpif recirc function.
- Fix use of return
This patch adds the necessary changes required to support
tunnel packet types in avx512 dpif.
Signed-off-by: Kumar Amber
Signed-off-by: Cian Ferriter
Co-authored-by: Cian Ferriter
---
v4:
- Rebase onto Simple match.
v3:
- Apply in_port optimization suggested by Harry.
---
---
Create new APIs for the avx512 DPIF, enabling one baseline
common code to be specialized into DPIF implementations for
"outer" processing, and "recirc" processing.
Signed-off-by: Kumar Amber
Signed-off-by: Cian Ferriter
Co-authored-by: Cian Ferriter
---
v4:
- Rebase onto master.
v3:
- Fix
The refactor allows us to use hash function accross
multiple files which was earlier restricted to
dpif-netdev.c only.
Signed-off-by: Kumar Amber
Signed-off-by: Cian Ferriter
Co-authored-by: Cian Ferriter
---
v3:
- Fix minor comments from Harry.
---
---
lib/dpif-netdev-private-dpcls.h | 23
The refactor allows us to use *recirc_depth_get() to obtain
the depth across ovs which was previously limited to only
dpif-netdev.c.
Signed-off-by: Kumar Amber
Signed-off-by: Cian Ferriter
Co-authored-by: Cian Ferriter
Acked-by: Harry van Haaren
---
lib/dpif-netdev-private-dpif.c | 2 ++
This Series of Patchsets introduce the Optimizations for
supporting tunneled packets in DPIF and MFEX. Along with
the optimization various refactoring of scalar
path is done to be used accross without duplication.
Over the Tests we have observed a gain of approximate 20~25%
gain in performance
On 8/22/2022 7:15 AM, David Marchand wrote:
On Mon, Aug 15, 2022 at 1:49 PM Ilya Maximets wrote:
On 8/8/22 22:36, Greg Rose wrote:
Remove the current xenserver implementation - it is obsolete and
since 3.0 we do not support kernel module builds [1].
1.
On Thu, Aug 25, 2022 at 1:57 PM Abhiram R N wrote:
>
> Hi Ihar,
>
> Thanks for the review of v4 patch.
> I will take care of the changes you have suggested in the xml(ovn-nb.xml),
> the tests (ovn-nbctl.at, ovn.at, ovn-northd.at) and ovn-nbctl.c in the next
> patch.
>
> Other than those
Hi Ihar,
Thanks for the review of v4 patch.
I will take care of the changes you have suggested in the xml(ovn-nb.xml),
the tests (ovn-nbctl.at, ovn.at, ovn-northd.at) and ovn-nbctl.c in the next
patch.
Other than those regarding the ovn-controller.c and ovn-nb.ovsschema I have
replied inline
On Thu, Aug 25, 2022 at 7:47 AM Xavier Simonart wrote:
>
> If a logical switch port is added and connected to a logical router
> port (through options: router-port) before the router port is
> created, then this might cause further issues such as segmentation
> violation when the switch and
On Thu, Aug 25, 2022 at 12:29:36PM +0200, Ilya Maximets wrote:
> On 8/24/22 21:29, Aaron Conole wrote:
> > lic121 writes:
> >
> >> Support running a list of scripts after vswitchd start, so that we can
> >> make some runtime config persistence across ovs restart. For example,
> >> the config
If a logical switch port is added and connected to a logical router
port (through options: router-port) before the router port is
created, then this might cause further issues such as segmentation
violation when the switch and router ports are deleted.
Signed-off-by: Xavier Simonart
---
v2:
A typical NVGRE encapsulated packet starts with the ETH/IP/GRE
protocols. Miniflow extract will parse just the ETH and IP headers. The
GRE header will be processed later as part of the pop action. Add
support for parsing the ETH/IP headers in this scenario.
Signed-off-by: Cian Ferriter
---
The subtable signatures being specialized here were found in an NVGRE
tunnel scenario.
Signed-off-by: Cian Ferriter
---
lib/dpif-netdev-lookup-avx512-gather.c | 4
lib/dpif-netdev-lookup-generic.c | 4
2 files changed, 8 insertions(+)
diff --git
This patchset adds AVX512 specialization for the NVGRE tunneling
scenario. It adds both an MFEX implementation for parsing the outer
header of a typical NVGRE packet and specialized DPCLS subtable
signatures for the lookups on both outer and inner headers.
Cian Ferriter (2):
dpif-netdev/dpcls:
Hello, David,
On Thu, Aug 25, 2022 at 2:52 PM David Marchand
wrote:
>
> Debian packaging target builds OVS against a packaged DPDK version.
> As a result, it is not relevant in the dpdk-latest branch which follows
> DPDK development branch.
>
> Signed-off-by: David Marchand
> ---
>
Debian packaging target builds OVS against a packaged DPDK version.
As a result, it is not relevant in the dpdk-latest branch which follows
DPDK development branch.
Signed-off-by: David Marchand
---
.github/workflows/build-and-test.yml | 1 +
1 file changed, 1 insertion(+)
diff --git
Hi,
On 17/08/2022 08:10, lic121 wrote:
In PMD, cycle cost for each rxq is saved so that we know each rxq's
load. But rxq that doesn't receive a packet is skipped. In fact,
polling no-packet rxq costs cycles as well. In my test, 100w pps
rxq(vhostuser) cost cycles 87,553,850,086 while no-packet
On 8/24/22 21:29, Aaron Conole wrote:
> lic121 writes:
>
>> Support running a list of scripts after vswitchd start, so that we can
>> make some runtime config persistence across ovs restart. For example,
>> the config that `ovs-appctl dpctl` makes.
>>
>> Signed-off-by: lic121
>> ---
>
> This
Propagating per_port_memory value through a DPDK netdev creation gives
the false impression its value is somehow contextual to the creation.
On the contrary, this parameter value is set once and for all at
OVS initialisation time.
Simplify the code and directly access the local boolean.
vhost related configuration and per port memory are netdev-dpdk
configuration items.
dpdk-stub.c and netdev-dpdk.c are never linked together, so we can move
those bits out of the generic dpdk code.
The dpdk_* accessors for those configuration items are then not needed
anymore and we can simply
Nothing earth-shattering in this series.
This is a followup after reviewing patches from Kevin: we can move
netdev-dpdk related configuration in this code and remove unneeded stubs
in the "dpdk" api in OVS (less code is always better :-)).
--
David Marchand
David Marchand (3):
netdev-dpdk:
Remove one unused stub: netdev_dpdk_register() can't be called if DPDK
is disabled at build time.
Remove unneeded #ifdef in call to free_dpdk_buf.
Drop unneeded cast when calling free_dpdk_buf.
Signed-off-by: David Marchand
---
lib/dp-packet.c | 6 +-
lib/dp-packet.h | 4 +---
active_v4_flows count was intialized when the northd node was computed.
However, neither sb_multicast_group nor en_sb_igmp_group causes northd
updates. Hence this count could keep increasing while processing igmp groups.
This issue was sometimes 'hidden' by northd recomputes due to lflow unable
If ovs-tcpdump received HUP or TERM signal, mirror and mirror
interface should be destroyed. This often happens, when
controlling terminal is closed, like ssh session closed, and
other users use kill to terminate it.
Signed-off-by: Daniel Ding
---
utilities/ovs-tcpdump.in | 30
Hi Han
Sorry for not replying earlier - I wanted to give a last chance for Dumitru
or others to comment
Thanks for applying the patch and all you feedback
Thanks
Xavier
On Thu, Aug 25, 2022 at 10:42 AM Dumitru Ceara wrote:
> On 8/25/22 08:35, Han Zhou wrote:
> > On Mon, Aug 22, 2022 at 9:41
On 8/24/22 18:32, Ilya Maximets wrote:
> On 8/24/22 17:57, Dumitru Ceara wrote:
>> On 8/24/22 17:49, Ilya Maximets wrote:
>>> On 8/24/22 17:43, Ilya Maximets wrote:
On 8/23/22 16:42, Dumitru Ceara wrote:
> After removing the possibility of disabling logical datapath groups the
> above
On 8/25/22 08:35, Han Zhou wrote:
> On Mon, Aug 22, 2022 at 9:41 AM Han Zhou wrote:
>>
>>
>>
>> On Fri, Aug 19, 2022 at 8:40 AM Xavier Simonart
> wrote:
>>>
>>> When VIF ports are claimed on a chassis, SBDB Port_Binding table is
> updated.
>>> If the SBDB IDL is still is read-only ("in
On Mon, Aug 22, 2022 at 9:41 AM Han Zhou wrote:
>
>
>
> On Fri, Aug 19, 2022 at 8:40 AM Xavier Simonart
wrote:
> >
> > When VIF ports are claimed on a chassis, SBDB Port_Binding table is
updated.
> > If the SBDB IDL is still is read-only ("in transaction") when such a
update
> > is required, the
On Tue, Aug 23, 2022 at 3:44 AM Xavier Simonart wrote:
>
> If a logical switch port is added and connected to a logical router
> port (through options: router-port) before the router port is
> created, then this might cause further issues such as segmentation
> violation when the switch and
35 matches
Mail list logo