Thanks for working on it!
Can we have nested clone?
actions=clone(push_vlan(1), resubmit(1,2), clone(mod_dl_src:,
output:5) )
BTW, I believe the port 3 will get a 104byte size packet with this action,
right?
actions= clone(truncate(100), push_vlan, output:3)
Thanks
Zhenyu Gao
2016-12-01 5:35
It looks better to this independent function netdev_linux_tap_batch_send.
Thanks for working on it.
Thanks
Zhenyu Gao
2017-08-03 5:59 GMT+08:00 Ben Pfaff :
> On Mon, Jul 17, 2017 at 02:39:57AM +, Zhenyu Gao wrote:
> > Sendmmsg can reduce cpu cycles in sending packets to
Thanks for working on it. LGTM.
Could you please add inline for function netdev_linux_tap_batch_send? That
makes it looks like function netdev_linux_sock_batch_send.
Or eliminate inline for function netdev_linux_sock_batch_send.
Thanks
Zhenyu Gao
2017-08-03 5:58 GMT+08:00 Ben Pfaff
Hi Loftus,
I had submitted a new version, please see
https://patchwork.ozlabs.org/patch/802070/
It move the cksum to vhost receive side.
Thanks
Zhenyu Gao
2017-08-10 12:35 GMT+08:00 Gao Zhenyu <sysugaozhe...@gmail.com>:
> I see, for flows in phy-phy setup, they should not be
I see, for flows in phy-phy setup, they should not be calculate cksum.
I will revise my patch to do the cksum for vhost port only. I will send a
new patch next week.
Thanks
Zhenyu Gao
2017-08-08 17:53 GMT+08:00 Loftus, Ciara :
> >
> > Hi Loftus,
> >
> > Thanks for
How about:
struct ovsdb_idl_table {
...
const struct ovsdb_idl_table_class *table_class
}
struct ovsdb_idl {
const struct ovsdb_idl_class *idl_class;
Besides of that, I see many places consume the table class.
Do you mind to make a macro helps to fetch the class?
Like:
How about mutable --> is_mutable ?
Thanks
Zhenyu Gao
2017-08-10 6:27 GMT+08:00 Joe Stringer :
> In C++, 'mutable' is a keyword. If this is used as the name for a field,
> then C++ compilers can get confused about the context and fail to
> compile references to such fields. Rename
us
tcp_lat:
latency = 82.3 us 83.3 us
tcp_lat:
latency = 93.1 us 97.2 us
tcp_lat:
latency = 237 us 211 us
2017-06-23 15:58 GMT+08:00 Chandran, Sugesh <sugesh.chand...@intel.com>:
>
>
>
>
> *Regards*
>
> *_Sugesh*
>
>
>
> *From:* Gao
Thanks for working on patch of implementing sendmmsg.
I will revise my patch base on it.
Thanks
Zhenyu Gao
2017-07-13 0:25 GMT+08:00 Ben Pfaff :
> On Fri, Jul 07, 2017 at 12:39:51AM +, Zhenyu Gao wrote:
> > Sendmmsg can reduce cpu cycles in sending packets to kernel.
> >
LGTM.
Thanks
Zhenyu Gao
2017-07-13 8:39 GMT+08:00 Ben Pfaff :
> This will have its first user in an upcoming commit.
>
> Signed-off-by: Ben Pfaff
> ---
> v1->v2: Note that FreeBSD also has sendmmsg. Fix sendmsg() return type.
>
> configure.ac| 3
Ping
Thanks
Zhenyu Gao
2017-07-17 10:39 GMT+08:00 Zhenyu Gao :
> Sendmmsg can reduce cpu cycles in sending packets to kernel.
> Replace sendmsg with sendmmsg in function netdev_linux_send to send
> batch packets if sendmmsg is available.
>
> If kernel side doesn't
for tcp/udp packets.
2017-07-20 10:52 GMT+08:00 Gao Zhenyu <sysugaozhe...@gmail.com>:
> Hi Sugesh,
>
> the setup like:
>
> qperf client
> +-+
> | VM|
> +-+
> |
> | qperf server
> +-
nts/questions below.
>
>
>
> *Regards*
>
> *_Sugesh*
>
>
>
> *From:* Gao Zhenyu [mailto:sysugaozhe...@gmail.com]
> *Sent:* Monday, July 17, 2017 12:55 PM
> *To:* Chandran, Sugesh <sugesh.chand...@intel.com>
> *Cc:* b...@ovn.org; u9012...@gmail.com; d...@open
ping
Thanks
Zhenyu Gao
2017-06-26 19:29 GMT+08:00 Zhenyu Gao :
> The element of rx_1024_to_1522_packets has wrong
> name(rx_1024_to_1518_packets).
> Change it from rx_1024_to_1518_packets to rx_1024_to_1522_packets, it
> should
> record packets between 1024 to 1522.
Hi Jarno,
Thanks for the suggestion, I will try to get a performance number and
submit it into Linux first. :)
Thanks
Zhenyu Gao
2017-04-22 4:24 GMT+08:00 Jarno Rajahalme :
> As a policy, Linux kernel datapath changes other than backports need to go
> to upstream Linux first,
ping
Thanks
Zhenyu Gao
2017-07-25 18:27 GMT+08:00 Zhenyu Gao :
> In dpdk_do_tx_copy function, all packets were copied to mbuf first,
> but QoS checking may drop some of them.
> Move the QoS checking in front of copying data to mbuf, it helps to
> reduce useless
Thanks for working on it!
I think new_ is not a good name. Could you please try to revise it?
like: old --> old_datum
new -->new_datum
BTW, you also need to update 'new' in the description of struct
ovsdb_idl_row which in the top of lib/ovsdb-idl-provider.h
Thanks
Zhenyu Gao
ek as I have a few concerns about the approach.
>
> Ian
>
> > -Original Message-
> > From: Darrell Ball [mailto:db...@vmware.com]
> > Sent: Tuesday, August 8, 2017 3:07 AM
> > To: Gao Zhenyu <sysugaozhe...@gmail.com>; Ben Pfaff <b...@ovn.org>;
&
LGTM
Thanks
Zhenyu Gao
2017-08-07 18:24 GMT+08:00 Christian Ehrhardt <
christian.ehrha...@canonical.com>:
> Ping, any update on including this?
>
> On Fri, Jul 14, 2017 at 8:36 AM, Christian Ehrhardt <
> christian.ehrha...@canonical.com> wrote:
>
> > If available use dpdk pkg-config info of
I think consuming a bit in reg to indicate if a packet is mirror packet or
regular packet is doable as well. And it is easy to implement.
Other table like port_security can check if it a mirror packet and do the
right actions(drop or forwarding to next table).
Thanks
Zhenyu Gao
2017-08-07 13:37
Hi Ian,
This patch is pending for a long time. May I know if I should revise it?
Thanks
Zhenyu Gao
2017-08-08 17:34 GMT+08:00 Gao Zhenyu <sysugaozhe...@gmail.com>:
> Thanks for the review. Please let me know if you have any concern on it. :)
>
> Thanks
> Zhenyu Gao
>
&
Please take a look at version 2: https://patchwork.ozlabs.org/patch/803731/
:)
Thanks
Zhenyu Gao
2017-08-18 15:15 GMT+08:00 Gao Zhenyu <sysugaozhe...@gmail.com>:
> Thanks for the comments.
> OK, I see. I will revise it and add testing in ovn tests, but it may be in
> next w
Thanks for the comments.
OK, I see. I will revise it and add testing in ovn tests, but it may be in
next week.
Thanks
Zhenyu Gao
2017-08-18 14:55 GMT+08:00 Miguel Angel Ajo Pelayo <majop...@redhat.com>:
>
>
> On Fri, Aug 18, 2017 at 3:24 AM, Gao Zhenyu <sysugaozhe...@
Hi IIya,
Thanks for working on it.
This patch consumes dp_packet_batch_clone so I have concern on the
performance. Could you please show some performace number with/without your
patch.
Thanks
Zhenyu Gao
2017-08-10 23:38 GMT+08:00 Ilya Maximets :
> While processing
Thanks for the suggestion. A testcase would be add in ovn testings. But I
am not familiar with ovn test and busy on other stuff now.
Since this issue affects many consumers who try to use HA gateways. I
prefer to push this fix in ovs master first, then we have more time to make
a testcase for it.
with vectorization?
Thanks
Zhenyu Gao
2017-06-21 16:03 GMT+08:00 Chandran, Sugesh <sugesh.chand...@intel.com>:
>
>
>
>
> *Regards*
>
> *_Sugesh*
>
>
>
> *From:* Gao Zhenyu [mailto:sysugaozhe...@gmail.com]
> *Sent:* Monday, June 19, 2017 1:23 PM
> *To:
* If 'may_steal' is false, the caller retains ownership of all the
* packets. If 'may_steal' is true, the caller transfers ownership of
all
* the packets to the network device, regardless of success.
*
* If 'concurrent_txq' is true, the caller may perform concurrent calls
ping
Zhenyu Gao
2017-06-20 18:15 GMT+08:00 Zhenyu Gao :
> Sendmmsg can reduce cpu cycles in sending packets to kernel.
> Replace sendmsg with sendmmsg in function netdev_linux_send to send
> batch packets if sendmmsg is available.
>
> If kernel side doesn't support
Hi Ben,
Thanks for catching it. I will add "--enable-Werror" next time.
BTW, I would like to submit another patch to use sendmmsg to replace
sendmsg. Sendmmsg get more benefit on throughput(my draft testing show 100%
improvment). Do you think it is doable?
Thanks
Zhenyu Gao
2017-05-31 0:43
-06-01 10:16 GMT+08:00 Ben Pfaff <b...@ovn.org>:
> Are you sure that this is the fastest way to interface OVS-DPDK to a
> container? But, even if it is not, optimizations are welcome.
>
> On Thu, Jun 01, 2017 at 09:50:44AM +0800, Gao Zhenyu wrote:
> > Here is the ba
ovs's bonding supports LAG, Active Backup, SLB.
http://docs.openvswitch.org/en/latest/topics/bonding/
2017-06-07 16:49 GMT+08:00 Sam :
> Hi all,
>
> For switch lag, there are some RFCs to define its behivour.
> For ovs bond port, is there some standard to define this?
>
> I
Thanks for that comments.
[Sugesh] Any reason, why this patch does only the TCP checksum offload??
The command line option says tx_checksum offload (it could be mistakenly
considered for full checksum offload).
[Zhenyu Gao] DPDK nic supports many hw offload feature like IPv4,IPV6,TCP,
Thansk for the questions.
the multipath_port can be set via ovn-nbctl.
Like : ovn-nbctl -- --id=@lrt create Logical_Router_Static_Route
ip_prefix=0.0.0.0/0 nexthop=10.88.77.1 multipath_port=[mp1,mp2] -- add
Logical_Router edge1 static_routes @lrt
This patch haven't implement a ovn-nbctl command
10:56 AM, Miguel Angel Ajo Pelayo
> <majop...@redhat.com> wrote:
> > On Thu, Sep 21, 2017 at 2:21 PM, Gao Zhenyu <sysugaozhe...@gmail.com>
> wrote:
> >
> >> I think the S/N or E/W are not the matter we should considering now.
> >>
> >> T
(from original chassis to destination chassis without going through the
> router chassis), such chassis was only used for N/S, but may be I got
> something wrong.
>
>
> On Wed, Sep 20, 2017 at 4:48 PM, Gao Zhenyu <sysugaozhe...@gmail.com>
> wrote:
>
>> "
>> But
t;> just a tunneled packet between chassis A and chassis B ?
>>
>> What's the benefit of multipath there if the possible failing link is
>> always the connection between chassis A and chassis B ?
>>
>> I suspect there's something I'm missing on the picture.
>>
>&
t if you could explain how would it work between several ports
> in the networks and routers (may be a diagram?) otherwise I can't be really
> helpful reviewing :)
>
> Cheers, and thanks for the patience.
>
> On Wed, Sep 20, 2017 at 12:25 PM, Gao Zhenyu <sysugaozhe...@gmail.com>
> wrote
k at lib/multipath.c lib/bundle.c you will
> find that bundle.c is going to consider the up/down status (slave_enabled
> check) of the links.
>
> That way the controller doesn't need to modify any flow based on link
> status.
>
> On Wed, Sep 20, 2017 at 5:45 AM, Gao Zhenyu &
Comments and suggestions are welcome :)
Thanks
Zhenyu Gao
2017-09-26 17:52 GMT+08:00 Zhenyu Gao :
> 1. ovn-nb.ovsschema was updated output_port field. Change the max entry
> number from 1 to unlimited.
> 2. Add multipath feature in ovn-northd part. northd generates
Hi llya,
Thanks for working it. Your patch tried to eliminate the may_steal in
dpdk qos, because may_steal handled on dpif-netdev layer and always true.
But in function dpdk_do_tx_copy, it set the may_steal to false, because
the packet buffer were not allocated by dpdk side so cannot released
Thanks for working on it.
I want to know if we can have a patch to try one more time if we hit the
hw_strip_crc error since we had introduce many options in ovs-dpdk now.
(Just like function dpdk_eth_dev_queue_setup, it figures out queue number
by retrying)
Thanks
Zhenyu Gao
2017-10-09 17:09
Thanks Derrell, it makes sence to me and I think the code path is more
clear.
Thanks
Zhenyu Gao
2017-09-06 0:24 GMT+08:00 Darrell Ball :
> Hi Gao
>
> How about the following incremental ?
>
> Thanks Darrell
>
>
>
> diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c
> index
Thanks for your testing and I reproduce it on my own machine.
I did the testing:
10% times to get about 8.5Gb/s thoughput when "ethtool -K eth0 tx on" , 90%
situation get 3.5Gb/s.
10% times to get about 3.5Gb/s thoughput when "ethtool -K eth0 tx off", 90%
situation I get 8.5Gb/s
And this wierd
;wang.qia...@zte.com.cn>> wrote:
>
> > I know your mean.
> > The receiver need to distinguish the traffic is regular or mirror. This
> > may need some special flow table to deal with it.
> >
> > Thanks
> >
> >
> >
> > *Gao Zhenyu <*sysug
anks,
> Billy.
>
>
> > -Original Message-
> > From: ovs-dev-boun...@openvswitch.org [mailto:ovs-dev-
> > boun...@openvswitch.org] On Behalf Of Gao Zhenyu
> > Sent: Wednesday, August 23, 2017 4:12 PM
> > To: Loftus, Ciara <ciara.lof...@intel.com>
> >
patch being discussed in this thread (also here:
> https://patchwork.ozlabs.org/patch/802070/) it seems does something very
> similar.
> Wondering on the feasibility of putting this functionality in a rte_vhost
> library call such that we don't have two separate implementations?
>
gt;
>
>
> Acked-by: ian.sto...@intel.com
>
>
>
> Ian
>
>
>
> *From:* Gao Zhenyu [mailto:sysugaozhe...@gmail.com]
> *Sent:* Thursday, August 31, 2017 2:56 AM
> *To:* Stokes, Ian <ian.sto...@intel.com>
> *Cc:* d...@openvswitch.org
> *Subject:* Re: [PAT
Hi Ciara,
Thanks for you comments!
I submit new version here with new testing result:
https://patchwork.ozlabs.org/patch/809042/:)
Thanks
Zhenyu Gao
2017-09-01 23:17 GMT+08:00 Loftus, Ciara :
> Hi Zhenyu,
>
> Thanks for the v3. No feedback yet on the common
e reviewing the patch this week and
> hope to have feedback by Friday for you. If you could rebase that would be
> helpful for testing.
>
>
>
> Thanks
>
> Ian
>
>
>
> *From:* Gao Zhenyu [mailto:sysugaozhe...@gmail.com]
> *Sent:* Friday, August 18, 2017 2:28 A
Here is the the testing I just did:
1. Add log in function netdev_dpdk_policer_pkt_handle
static inline
bool
netdev_dpdk_policer_pkt_handle(struct rte_meter_srtcm
*meter,
struct rte_mbuf *pkt, uint64_t
time)
{
uint32_t pkt_len = rte_pktmbuf_pkt_len(pkt) -
s in this batch, may hit issue I think
}
2017-10-09 14:16 GMT+08:00 Ilya Maximets <i.maxim...@samsung.com>:
> On 08.10.2017 12:32, Gao Zhenyu wrote:
> > Hi llya,
> >
> > Thanks for working it. Your patch tried to eliminate the may_steal in
> dpdk qos, because may_
on the mailing list, or do you want me to do it?
Best regards,
=
2017-10-08 17:42 GMT+08:00 Gao Zhenyu <sysugaozhe...@gmail.com>:
> Comments and suggestions are welcome :)
>
> Thanks
> Zhenyu Gao
>
> 2017-09-26 17:52 GMT+08:00 Zheny
y get a idea to make it, it would be great to bring it
up then we can discuss it and move whole process faster. :)
Thanks
Zhenyu Gao
2017-10-11 9:50 GMT+08:00 Gao Zhenyu <sysugaozhe...@gmail.com>:
> I discussed this multipath stuff with Miguel in other mailling thread and
> I want
ping.
Thanks
Zhenyu Gao
2017-10-11 9:55 GMT+08:00 Gao Zhenyu <sysugaozhe...@gmail.com>:
> Hi Miguel,
>
>Thanks for your suggestion on it. It's very usefull.
>In my point of view, I think no matter we have single router leg or
> multiple router legs on edge
Hi Ben,
Thanks for the comments!
The multipath feature can separete traffics to multi-gateways. If you
get two gateway routers and both of them pin on chassis. (gateway1 pin on
chassis-A, gateway2 pin on chassis-B) The multipath can seperate traffic to
those chassis automaticly.
Thanks for the review! The swapping may break it. I would try to revise it.
Thanks
Zhenyu Gao
2018-01-09 3:10 GMT+08:00 Ben Pfaff :
> On Mon, Jan 08, 2018 at 11:36:22AM +, Zhenyu Gao wrote:
> > 1. ovs cannot create two ports with same tunnel type and
> options:remote_ip
> > 2.
Hi guys,
As you know ovn consumes rconn functions in pinctrl.c to connect the
bridge. But the rconn functions are private functions. I mean those
functions are not exposed in include/openvswitch/*.h file, so they are not
regular APIs, right?
So do you know if we get a way to connect
Thanks for the suggestion. I will try to make them become regular APIs and
submit patch later.
Thanks
Zhenyu Gao
2018-01-03 0:37 GMT+08:00 Ben Pfaff <b...@ovn.org>:
> On Fri, Dec 29, 2017 at 11:39:16AM +0800, Gao Zhenyu wrote:
> > As you know ovn consumes rconn functio
sorry. I just realize the rconn.h had become a public APIs a month ago :)
Thanks
Zhenyu Gao
2018-01-03 9:21 GMT+08:00 Gao Zhenyu <sysugaozhe...@gmail.com>:
> Thanks for the suggestion. I will try to make them become regular APIs and
> submit patch later.
>
> Thanks
> Zhen
Hi
Found a potential memleak issue when I enable bridge ipfix and transmit
traffic through geneve tunnel.
The kernel is centos7.5 3.10.0-862.9.1.el7.x86_64
memleak show a lot of malloc call-stack :
unreferenced object 0x93f473e64000 (size 512):
comm "softirq", pid 0, jiffies
60 matches
Mail list logo