Re: [vpp-dev] The VCL server CPU utilization is 100% ， if there is no message to the epoll_wait

[xyxue]

Hi daw,

Thank you for your reply . Another question :If the 'listen ,send,recv,connect' 
will cause the cpu utilization 100%?

Thanks,
Xyxue


 
发件人： Dave Wallace
发送时间： 2018-06-05 02:22
收件人： vpp-dev; 薛欣颖
主题： Re: [vpp-dev] The VCL server CPU utilization is 100% ， if there is no 
message to the epoll_wait
Xyxue,

This is a known by-product of the existing [prototype] epoll_wait 
implementation. There is currently no mechanism in epoll_wait to have the 
thread block on a condvar which would cause the thread to sleep instead of 
sitting in a continuous polling loop.

Fixing this is on the list of things to do in VCL, but there is currently no 
ETA for when it will be resolved.

Thanks,
-daw-

On 6/4/2018 6:13 AM, xyxue wrote:

Hi guys，

I‘m testing the VCL . The VCL server CPU utilization is 100% ,if there is no 
message to the epoll_wait. Is there anything I can do to slove it?
Is there  the same problem in the VCL client?

Thanks,
Xyxue

Re: [vpp-dev] [FD.io Helpdesk #56625] Nexus fd.io.master.centos7 VPP artifacts

[Peter Mikus via Lists.Fd.Io]

Hello Vanessa,

Thank you for an explanation. Indeed this will impact certain things that are 
planned like "automatic bisecting" (downloading and testing range of 
artifacts). Let me analyze current situation with CSIT team and get back to you.

Peter Mikus
Engineer – Software
Cisco Systems Limited


-Original Message-
From: Vanessa Valderrama via RT [mailto:fdio-helpd...@rt.linuxfoundation.org] 
Sent: Monday, June 04, 2018 9:47 PM
To: Peter Mikus -X (pmikus - PANTHEON TECHNOLOGIES at Cisco) 
Cc: csit-...@lists.fd.io; infra-steer...@lists.fd.io; vpp-dev@lists.fd.io
Subject: [FD.io Helpdesk #56625] Nexus fd.io.master.centos7 VPP artifacts

Peter,

The fd.io.master.centos7 repo had to be cleaned up significantly to eliminate 
Jenkins build timeout errors.  This was discussed in the TSC. Going forward 
we'll only be keeping an average of 10 of the current release candidate 
artifacts in the repository.  Please let me know if this retention policy 
causes an issue for you.

We do need to clean up the other repositories as well.  Please let me know if 
you'd like to discuss retention policies.  I'll hold off on cleaning up other 
repositories for now.

Thank you,
Vanessa

On Wed May 30 10:20:21 2018, pmi...@cisco.com wrote:
> Hello,
> 
> I have recently spotted that CentOS repo got reduced and old binaries 
> are missing [1].
> 
> Is this expected?
> Will the similar be done for Ubuntu repos?
> 
> Was this announced somewhere?
> 
> Thank you.
> 
> [1]
> https://nexus.fd.io/content/repositories/fd.io.master.centos7/io/fd/vp
> p/vpp/
> 
> Peter Mikus
> Engineer - Software
> Cisco Systems Limited
> [http://www.cisco.com/web/europe/images/email/signature/logo05.jpg]
> Think before you print.
> This email may contain confidential and privileged material for the 
> sole use of the intended recipient. Any review, use, distribution or 
> disclosure by others is strictly prohibited. If you are not the 
> intended recipient (or authorized to receive for the recipient), 
> please contact the sender by reply email and delete all copies of this 
> message.
> For corporate legal information go to:
> http://www.cisco.com/web/about/doing_business/legal/cri/index.html




-=-=-=-=-=-=-=-=-=-=-=-
Links:

You receive all messages sent to this group.

View/Reply Online (#9525): https://lists.fd.io/g/vpp-dev/message/9525
View All Messages In Topic (2): https://lists.fd.io/g/vpp-dev/topic/21275985
Mute This Topic: https://lists.fd.io/mt/21275985/21656
New Topic: https://lists.fd.io/g/vpp-dev/post

Change Your Subscription: https://lists.fd.io/g/vpp-dev/editsub/21656
Group Home: https://lists.fd.io/g/vpp-dev
Contact Group Owner: vpp-dev+ow...@lists.fd.io
Terms of Service: https://lists.fd.io/static/tos
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub
Email sent to: arch...@mail-archive.com
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] VPP Vnet crash with vhost-user interface

[Ravi Kerur]

Hi Steven,

Thanks for your help. I am using vhost-user client (VPP in container)
and vhost-user server (VPP in host). I thought it should work.

create vhost socket /var/run/vpp/sock3.sock server (On host)

create vhost socket /var/run/usvhost1 (On container)

Can you please point me to a document which shows how to create VPP
virtio_user interfaces or static configuration in
/etc/vpp/startup.conf?

I have used following declarations in /etc/vpp/startup.conf

# vdev virtio_user0,path=/var/run/vpp/sock3.sock,mac=52:54:00:00:04:01
# vdev virtio_user1,path=/var/run/vpp/sock4.sock,mac=52:54:00:00:04:02

but it doesn't work.

Thanks.

On Mon, Jun 4, 2018 at 3:57 PM, Steven Luong (sluong)  wrote:
> Ravi,
>
> VPP only supports vhost-user in the device mode. In your example, the host, 
> in device mode, and the container also in device mode do not make a happy 
> couple. You need one of them, either the host or container, running in driver 
> mode using the dpdk vdev virtio_user command in startup.conf. So you need 
> something like this
>
> (host) VPP native vhost-user - (container) VPP DPDK vdev virtio_user
>   -- or --
> (host) VPP DPDK vdev virtio_user  (container) VPP native vhost-user
>
> Steven
>
> On 6/4/18, 3:27 PM, "Ravi Kerur"  wrote:
>
> Hi Steven
>
> Though crash is not happening anymore, there is still an issue with Rx
> and Tx. To eliminate whether it is testpmd or vpp, I decided to run
>
> (1) VPP vhost-user server on host-x
> (2) Run VPP in a container on host-x and vhost-user client port
> connecting to vhost-user server.
>
> Still doesn't work. Details below. Please let me know if something is
> wrong in what I am doing.
>
>
> (1) VPP vhost-user as a server
> (2) VPP in a container virtio-user or vhost-user client
>
> (1) Create vhost-user server socket on VPP running on host.
>
> vpp#create vhost socket /var/run/vpp/sock3.sock server
> vpp#set interface state VirtualEthernet0/0/0 up
> show vhost-user VirtualEthernet0/0/0 descriptors
> Virtio vhost-user interfaces
> Global:
> coalesce frames 32 time 1e-3
> number of rx virtqueues in interrupt mode: 0
> Interface: VirtualEthernet0/0/0 (ifindex 3)
> virtio_net_hdr_sz 0
> features mask (0x):
> features (0x0):
> protocol features (0x0)
>
> socket filename /var/run/vpp/sock3.sock type server errno "Success"
>
> rx placement:
> tx placement: spin-lock
> thread 0 on vring 0
>
> Memory regions (total 0)
>
> vpp# set interface ip address VirtualEthernet0/0/0 192.168.1.1/24
> vpp#
>
> (2) Instantiate a docker container to run VPP connecting to sock3.server 
> socket.
>
> docker run -it --privileged -v
> /var/run/vpp/sock3.sock:/var/run/usvhost1 -v
> /dev/hugepages:/dev/hugepages dpdk-app-vpp:latest
> root@4b1bd06a3225:~/dpdk#
> root@4b1bd06a3225:~/dpdk# ps -ef
> UID PID PPID C STIME TTY TIME CMD
> root 1 0 0 21:39 ? 00:00:00 /bin/bash
> root 17 1 0 21:39 ? 00:00:00 ps -ef
> root@4b1bd06a3225:~/dpdk#
>
> root@8efda6701ace:~/dpdk# ps -ef | grep vpp
> root 19 1 39 21:41 ? 00:00:03 /usr/bin/vpp -c /etc/vpp/startup.conf
> root 25 1 0 21:41 ? 00:00:00 grep --color=auto vpp
> root@8efda6701ace:~/dpdk#
>
> vpp#create vhost socket /var/run/usvhost1
> vpp#set interface state VirtualEthernet0/0/0 up
> vpp#show vhost-user VirtualEthernet0/0/0 descriptors
> Virtio vhost-user interfaces
> Global:
> coalesce frames 32 time 1e-3
> number of rx virtqueues in interrupt mode: 0
> Interface: VirtualEthernet0/0/0 (ifindex 1)
> virtio_net_hdr_sz 0
> features mask (0x):
> features (0x0):
> protocol features (0x0)
>
> socket filename /var/run/usvhost1 type client errno "Success"
>
> rx placement:
> tx placement: spin-lock
> thread 0 on vring 0
>
> Memory regions (total 0)
>
> vpp#
>
> vpp# set interface ip address VirtualEthernet0/0/0 192.168.1.2/24
> vpp#
>
> vpp# ping 192.168.1.1
>
> Statistics: 5 sent, 0 received, 100% packet loss
> vpp#
>
> On Thu, May 31, 2018 at 2:30 PM, Steven Luong (sluong)  
> wrote:
> > show interface and look for the counter and count columns for the 
> corresponding interface.
> >
> > Steven
> >
> > On 5/31/18, 1:28 PM, "Ravi Kerur"  wrote:
> >
> > Hi Steven,
> >
> > You made my day, thank you. I didn't realize different dpdk versions
> > (vpp -- 18.02.1 and testpmd -- from latest git repo (probably 18.05)
> > could be the cause of the problem, I still dont understand why it
> > should as virtio/vhost messages are meant to setup tx/rx rings
> > correctly?
> >
> > I downloaded dpdk 18.02.1 stable release and at least vpp doesn't
> > crash now (for both vpp-native and dpdk vhost interfaces). I have 
> one
> > question 

Re: [vpp-dev] VPP Vnet crash with vhost-user interface

[steven luong via Lists.Fd.Io]

Ravi,

VPP only supports vhost-user in the device mode. In your example, the host, in 
device mode, and the container also in device mode do not make a happy couple. 
You need one of them, either the host or container, running in driver mode 
using the dpdk vdev virtio_user command in startup.conf. So you need something 
like this

(host) VPP native vhost-user - (container) VPP DPDK vdev virtio_user
  -- or --
(host) VPP DPDK vdev virtio_user  (container) VPP native vhost-user

Steven

On 6/4/18, 3:27 PM, "Ravi Kerur"  wrote:

Hi Steven

Though crash is not happening anymore, there is still an issue with Rx
and Tx. To eliminate whether it is testpmd or vpp, I decided to run

(1) VPP vhost-user server on host-x
(2) Run VPP in a container on host-x and vhost-user client port
connecting to vhost-user server.

Still doesn't work. Details below. Please let me know if something is
wrong in what I am doing.


(1) VPP vhost-user as a server
(2) VPP in a container virtio-user or vhost-user client

(1) Create vhost-user server socket on VPP running on host.

vpp#create vhost socket /var/run/vpp/sock3.sock server
vpp#set interface state VirtualEthernet0/0/0 up
show vhost-user VirtualEthernet0/0/0 descriptors
Virtio vhost-user interfaces
Global:
coalesce frames 32 time 1e-3
number of rx virtqueues in interrupt mode: 0
Interface: VirtualEthernet0/0/0 (ifindex 3)
virtio_net_hdr_sz 0
features mask (0x):
features (0x0):
protocol features (0x0)

socket filename /var/run/vpp/sock3.sock type server errno "Success"

rx placement:
tx placement: spin-lock
thread 0 on vring 0

Memory regions (total 0)

vpp# set interface ip address VirtualEthernet0/0/0 192.168.1.1/24
vpp#

(2) Instantiate a docker container to run VPP connecting to sock3.server 
socket.

docker run -it --privileged -v
/var/run/vpp/sock3.sock:/var/run/usvhost1 -v
/dev/hugepages:/dev/hugepages dpdk-app-vpp:latest
root@4b1bd06a3225:~/dpdk#
root@4b1bd06a3225:~/dpdk# ps -ef
UID PID PPID C STIME TTY TIME CMD
root 1 0 0 21:39 ? 00:00:00 /bin/bash
root 17 1 0 21:39 ? 00:00:00 ps -ef
root@4b1bd06a3225:~/dpdk#

root@8efda6701ace:~/dpdk# ps -ef | grep vpp
root 19 1 39 21:41 ? 00:00:03 /usr/bin/vpp -c /etc/vpp/startup.conf
root 25 1 0 21:41 ? 00:00:00 grep --color=auto vpp
root@8efda6701ace:~/dpdk#

vpp#create vhost socket /var/run/usvhost1
vpp#set interface state VirtualEthernet0/0/0 up
vpp#show vhost-user VirtualEthernet0/0/0 descriptors
Virtio vhost-user interfaces
Global:
coalesce frames 32 time 1e-3
number of rx virtqueues in interrupt mode: 0
Interface: VirtualEthernet0/0/0 (ifindex 1)
virtio_net_hdr_sz 0
features mask (0x):
features (0x0):
protocol features (0x0)

socket filename /var/run/usvhost1 type client errno "Success"

rx placement:
tx placement: spin-lock
thread 0 on vring 0

Memory regions (total 0)

vpp#

vpp# set interface ip address VirtualEthernet0/0/0 192.168.1.2/24
vpp#

vpp# ping 192.168.1.1

Statistics: 5 sent, 0 received, 100% packet loss
vpp#

On Thu, May 31, 2018 at 2:30 PM, Steven Luong (sluong)  
wrote:
> show interface and look for the counter and count columns for the 
corresponding interface.
>
> Steven
>
> On 5/31/18, 1:28 PM, "Ravi Kerur"  wrote:
>
> Hi Steven,
>
> You made my day, thank you. I didn't realize different dpdk versions
> (vpp -- 18.02.1 and testpmd -- from latest git repo (probably 18.05)
> could be the cause of the problem, I still dont understand why it
> should as virtio/vhost messages are meant to setup tx/rx rings
> correctly?
>
> I downloaded dpdk 18.02.1 stable release and at least vpp doesn't
> crash now (for both vpp-native and dpdk vhost interfaces). I have one
> question is there a way to read vhost-user statistics counter (Rx/Tx)
> on vpp? I only know
>
> 'show vhost-user ' and 'show vhost-user  descriptors'
> which doesn't show any counters.
>
> Thanks.
>
> On Thu, May 31, 2018 at 11:51 AM, Steven Luong (sluong)
>  wrote:
> > Ravi,
> >
> > For (1) which works, what dpdk version are you using in the host? 
Are you using the same dpdk version as VPP is using? Since you are using VPP 
latest, I think it is 18.02. Type "show dpdk version" at the VPP prompt to find 
out for sure.
> >
> > Steven
> >
> > On 5/31/18, 11:44 AM, "Ravi Kerur"  wrote:
> >
> > Hi Steven,
> >
> > i have tested following scenarios and it basically is not clear 
why

Re: [vpp-dev] VPP Vnet crash with vhost-user interface

[Ravi Kerur]

Hi Steven

Though crash is not happening anymore, there is still an issue with Rx
and Tx. To eliminate whether it is testpmd or vpp, I decided to run

(1) VPP vhost-user server on host-x
(2) Run VPP in a container on host-x and vhost-user client port
connecting to vhost-user server.

Still doesn't work. Details below. Please let me know if something is
wrong in what I am doing.


(1) VPP vhost-user as a server
(2) VPP in a container virtio-user or vhost-user client

(1) Create vhost-user server socket on VPP running on host.

vpp#create vhost socket /var/run/vpp/sock3.sock server
vpp#set interface state VirtualEthernet0/0/0 up
show vhost-user VirtualEthernet0/0/0 descriptors
Virtio vhost-user interfaces
Global:
coalesce frames 32 time 1e-3
number of rx virtqueues in interrupt mode: 0
Interface: VirtualEthernet0/0/0 (ifindex 3)
virtio_net_hdr_sz 0
features mask (0x):
features (0x0):
protocol features (0x0)

socket filename /var/run/vpp/sock3.sock type server errno "Success"

rx placement:
tx placement: spin-lock
thread 0 on vring 0

Memory regions (total 0)

vpp# set interface ip address VirtualEthernet0/0/0 192.168.1.1/24
vpp#

(2) Instantiate a docker container to run VPP connecting to sock3.server socket.

docker run -it --privileged -v
/var/run/vpp/sock3.sock:/var/run/usvhost1 -v
/dev/hugepages:/dev/hugepages dpdk-app-vpp:latest
root@4b1bd06a3225:~/dpdk#
root@4b1bd06a3225:~/dpdk# ps -ef
UID PID PPID C STIME TTY TIME CMD
root 1 0 0 21:39 ? 00:00:00 /bin/bash
root 17 1 0 21:39 ? 00:00:00 ps -ef
root@4b1bd06a3225:~/dpdk#

root@8efda6701ace:~/dpdk# ps -ef | grep vpp
root 19 1 39 21:41 ? 00:00:03 /usr/bin/vpp -c /etc/vpp/startup.conf
root 25 1 0 21:41 ? 00:00:00 grep --color=auto vpp
root@8efda6701ace:~/dpdk#

vpp#create vhost socket /var/run/usvhost1
vpp#set interface state VirtualEthernet0/0/0 up
vpp#show vhost-user VirtualEthernet0/0/0 descriptors
Virtio vhost-user interfaces
Global:
coalesce frames 32 time 1e-3
number of rx virtqueues in interrupt mode: 0
Interface: VirtualEthernet0/0/0 (ifindex 1)
virtio_net_hdr_sz 0
features mask (0x):
features (0x0):
protocol features (0x0)

socket filename /var/run/usvhost1 type client errno "Success"

rx placement:
tx placement: spin-lock
thread 0 on vring 0

Memory regions (total 0)

vpp#

vpp# set interface ip address VirtualEthernet0/0/0 192.168.1.2/24
vpp#

vpp# ping 192.168.1.1

Statistics: 5 sent, 0 received, 100% packet loss
vpp#

On Thu, May 31, 2018 at 2:30 PM, Steven Luong (sluong)  wrote:
> show interface and look for the counter and count columns for the 
> corresponding interface.
>
> Steven
>
> On 5/31/18, 1:28 PM, "Ravi Kerur"  wrote:
>
> Hi Steven,
>
> You made my day, thank you. I didn't realize different dpdk versions
> (vpp -- 18.02.1 and testpmd -- from latest git repo (probably 18.05)
> could be the cause of the problem, I still dont understand why it
> should as virtio/vhost messages are meant to setup tx/rx rings
> correctly?
>
> I downloaded dpdk 18.02.1 stable release and at least vpp doesn't
> crash now (for both vpp-native and dpdk vhost interfaces). I have one
> question is there a way to read vhost-user statistics counter (Rx/Tx)
> on vpp? I only know
>
> 'show vhost-user ' and 'show vhost-user  descriptors'
> which doesn't show any counters.
>
> Thanks.
>
> On Thu, May 31, 2018 at 11:51 AM, Steven Luong (sluong)
>  wrote:
> > Ravi,
> >
> > For (1) which works, what dpdk version are you using in the host? Are 
> you using the same dpdk version as VPP is using? Since you are using VPP 
> latest, I think it is 18.02. Type "show dpdk version" at the VPP prompt to 
> find out for sure.
> >
> > Steven
> >
> > On 5/31/18, 11:44 AM, "Ravi Kerur"  wrote:
> >
> > Hi Steven,
> >
> > i have tested following scenarios and it basically is not clear why
> > you think DPDK is the problem? Is it possible VPP and DPDK use
> > different virtio versions?
> >
> > Following are the scenarios I have tested
> >
> > (1) testpmd/DPDK vhost-user (running on host) and testpmd/DPDK
> > virito-user (in a container) -- can send and receive packets
> > (2) VPP-native vhost-user (running on host) and testpmd/DPDK
> > virtio-user (in a container) -- VPP crashes and it is in VPP code
> > (3) VPP-DPDK vhost user (running on host) and testpmd/DPDK 
> virtio-user
> > (in a container) -- VPP crashes and in DPDK
> >
> > Thanks.
> >
> > On Thu, May 31, 2018 at 10:12 AM, Steven Luong (sluong)
> >  wrote:
> > > Ravi,
> > >
> > > I've proved my point -- there is a problem in the way that you 
> invoke testpmd. The shared memory region that it passes to the device is not 
> accessible from the device. I don't know what the correct options are that 
> you need to use. This is really a question for dpdk.
> 

[vpp-dev] [FD.io Helpdesk #56625] Nexus fd.io.master.centos7 VPP artifacts

[Vanessa Valderrama via RT]

Peter,

The fd.io.master.centos7 repo had to be cleaned up significantly to eliminate 
Jenkins build timeout errors.  This was discussed in the TSC. Going forward 
we'll only be keeping an average of 10 of the current release candidate 
artifacts in the repository.  Please let me know if this retention policy 
causes an issue for you.

We do need to clean up the other repositories as well.  Please let me know if 
you'd like to discuss retention policies.  I'll hold off on cleaning up other 
repositories for now.

Thank you,
Vanessa

On Wed May 30 10:20:21 2018, pmi...@cisco.com wrote:
> Hello,
> 
> I have recently spotted that CentOS repo got reduced and old binaries
> are missing [1].
> 
> Is this expected?
> Will the similar be done for Ubuntu repos?
> 
> Was this announced somewhere?
> 
> Thank you.
> 
> [1]
> https://nexus.fd.io/content/repositories/fd.io.master.centos7/io/fd/vpp/vpp/
> 
> Peter Mikus
> Engineer - Software
> Cisco Systems Limited
> [http://www.cisco.com/web/europe/images/email/signature/logo05.jpg]
> Think before you print.
> This email may contain confidential and privileged material for the
> sole use of the intended recipient. Any review, use, distribution or
> disclosure by others is strictly prohibited. If you are not the
> intended recipient (or authorized to receive for the recipient),
> please contact the sender by reply email and delete all copies of this
> message.
> For corporate legal information go to:
> http://www.cisco.com/web/about/doing_business/legal/cri/index.html




-=-=-=-=-=-=-=-=-=-=-=-
Links:

You receive all messages sent to this group.

View/Reply Online (#9521): https://lists.fd.io/g/vpp-dev/message/9521
View All Messages In Topic (1): https://lists.fd.io/g/vpp-dev/topic/21275985
Mute This Topic: https://lists.fd.io/mt/21275985/21656
New Topic: https://lists.fd.io/g/vpp-dev/post

Change Your Subscription: https://lists.fd.io/g/vpp-dev/editsub/21656
Group Home: https://lists.fd.io/g/vpp-dev
Contact Group Owner: vpp-dev+ow...@lists.fd.io
Terms of Service: https://lists.fd.io/static/tos
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub
Email sent to: arch...@mail-archive.com
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] The VCL server CPU utilization is 100% ， if there is no message to the epoll_wait

[Dave Wallace]

Xyxue,

This is a known by-product of the existing [prototype] epoll_wait 
implementation. There is currently no mechanism in epoll_wait to have 
the thread block on a condvar which would cause the thread to sleep 
instead of sitting in a continuous polling loop.


Fixing this is on the list of things to do in VCL, but there is 
currently no ETA for when it will be resolved.


Thanks,
-daw-

On 6/4/2018 6:13 AM, xyxue wrote:


Hi guys，

I‘m testing the VCL . The VCL server CPU utilization is 100% ,if there 
is no message to the epoll_wait. Is there anything I can do to slove it?

Is there  the same problem in the VCL client?

Thanks,
Xyxue

Re: [vpp-dev] dpdk shared libraries

[Thomas F Herbert]

On 06/04/2018 09:28 AM, Damjan Marion (damarion) wrote:


Dear Thomas,

That should be doable today, as long as VPP version supports version 
of dpdk you want to link against.


From the spec file you can invoke autotools project directly, and pass 
--enable-dpdk-shared to ./configure


./configure --enable-dpdk-shared 
--prefix=/wherever/rpm/spec/file/wants/files/to/show/up

make
make install

I don't expect any perf difference, but the open question is how the 
distro dpdk libs deal with different vector optimisations

I assume you are speaking of AVX?

It looks like upstream DPDK (default settings) and the fedora RPM both 
have AVX2 enabled but not AVX512.

I think that is consistent of what we do in VPP when we compile dpdk.


--
Damjan

On 4 Jun 2018, at 13:51, Thomas F Herbert > wrote:


OVS as of 9.x now uses shared libraries with dpdk packaged separately 
for downstream


I want to raise the issue of shared vs. static libraries for dpdk in vpp.

Would there be any performance penalty to going with shared?

--Tom

--
*Thomas F Herbert*
NFV and Fast Data Planes
Networking Group Office of the CTO
*Red Hat*





--
*Thomas F Herbert*
NFV and Fast Data Planes
Networking Group Office of the CTO
*Red Hat*

[vpp-dev] gerrit job failing on ARM due to pre-installed dpdk package

[Nitin Saxena]

Hi,

Gerrit patch: https://gerrit.fd.io/r/#/c/12813/ failing because dpdk 
package is pre-installed on ThunderX and not compiled during this patch.


Logs from: 
https://jenkins.fd.io/job/vpp-arm-verify-master-ubuntu1604/617/console

=
12:22:26 make -C dpdk install-deb
12:22:26 make[1]: Entering directory 
'/w/workspace/vpp-arm-verify-master-ubuntu1604/dpdk'

12:22:26 ==
12:22:26  Up-to-date DPDK package already installed
12:22:26 ==
12:22:26 make[1]: Leaving directory 
'/w/workspace/vpp-arm-verify-master-ubuntu1604/dpdk

=

This patch require dpdk package to re-compile and install again. How to 
trigger it on ThunderX?


Thanks,
Nitin


-=-=-=-=-=-=-=-=-=-=-=-
Links:

You receive all messages sent to this group.

View/Reply Online (#9518): https://lists.fd.io/g/vpp-dev/message/9518
View All Messages In Topic (1): https://lists.fd.io/g/vpp-dev/topic/21233595
Mute This Topic: https://lists.fd.io/mt/21233595/21656
New Topic: https://lists.fd.io/g/vpp-dev/post

Change Your Subscription: https://lists.fd.io/g/vpp-dev/editsub/21656
Group Home: https://lists.fd.io/g/vpp-dev
Contact Group Owner: vpp-dev+ow...@lists.fd.io
Terms of Service: https://lists.fd.io/static/tos
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub
Email sent to: arch...@mail-archive.com
-=-=-=-=-=-=-=-=-=-=-=-

Re: [vpp-dev] Query on mac learning - VLAN

[John Lo (loj)]

On VPP, packets with VLANs does not imply it go into a separate bridge domains 
(BDs).  For your case, your sub-interfaces for VLAN 100 and 200 on your host 
interface should be put into separate bridge domains, e.g. you can use BD IDs 
100 and 200, and have the correct behavior.  As I described in an earlier email 
thread on this mailing list, all packets in a BD are expected to have their 
first VLAN tag with the same ID.  You currently have both sub-interfaces in the 
same BD so is not the correct setup. Your "expected output" example is not 
correct as the MACs on different sub-interfaces should be in different BDs thus 
different BD-Idx.

DBGvpp# set int l2 bridge ?
  set interface l2 bridge  set interface l2 bridge  
 [bvi] [shg]
DBGvpp# show bridge-domain ?
  show bridge-domain   show bridge-domain [bridge-domain-id 
[detail|int|arp|bd-tag]]

Regards,
John

From: vpp-dev@lists.fd.io  On Behalf Of Subramanya Bhatkal 
Nagesh
Sent: Monday, June 04, 2018 8:00 AM
To: vpp-dev@lists.fd.io
Cc: Padmavathy Ramakrishnan 
Subject: [vpp-dev] Query on mac learning - VLAN

Hi,

We are interested in understanding if we can learn mac address on a per VLAN 
basis in VPP.

If an interface 'X' is part of VLAN 100, and interface 'Y' is part of VLAN 200
A host 'H' with mac "00:10:94:00:00:33" is behind a switch, ports of which are 
part of both VLAN 100 and VLAN 200.
Mac should be learnt for host 'H' on both VLAN's 100 and 200.

Traffic shall be forwarded to host based on the VLAN.

In the current implementation, by default the mac is learnt on the last learnt 
interface

DBGvpp# show l2fib verbose
Mac-Address BD-Idx If-Idx BSN-ISN Age(min) static filter bvi 
Interface-Name
 08:00:27:aa:bb:211  4  0/1  -   -  - - 
GigabitEthernet0/8/0.100
L2FIB total/learned entries: 1/1  Last scan time: 0.e0sec  Learn limit: 
4194304

DBGvpp# l2fib add 08:00:27:aa:bb:21 100 GigabitEthernet0/9/0.100

DBGvpp# show l2fib verbose
Mac-Address BD-Idx If-Idx BSN-ISN Age(min) static filter bvi 
Interface-Name
 08:00:27:aa:bb:211  3  0/0  -   -  - - 
GigabitEthernet0/9/0.100
L2FIB total/learned entries: 1/0  Last scan time: 0.e0sec  Learn limit: 
4194304

Expected Table:
Table shall look something like below
Mac-Address BD-Idx If-Idx BSN-ISN Age(min) static filter bvi 
Interface-Name
 00:10:94:00:00:331  4  0/0  -   -  - - 
GigabitEthernet0/8/0.100
00:10:94:00:00:331  3  0/0  -   -  - - 
GigabitEthernet0/9/0.200
L2FIB total/learned entries: 2/2  Last scan time: 0.e0sec  Learn limit: 
4194304


Regards,
Subramanya
Radisys

Re: [vpp-dev] dpdk shared libraries

[Damjan Marion]

Dear Thomas,

That should be doable today, as long as VPP version supports version of dpdk 
you want to link against.

>From the spec file you can invoke autotools project directly, and pass 
>--enable-dpdk-shared to ./configure

./configure --enable-dpdk-shared 
--prefix=/wherever/rpm/spec/file/wants/files/to/show/up
make
make install

I don't expect any perf difference, but the open question is how the distro 
dpdk libs deal with different vector optimisations

--
Damjan

On 4 Jun 2018, at 13:51, Thomas F Herbert 
mailto:therb...@redhat.com>> wrote:


OVS as of 9.x now uses shared libraries with dpdk packaged separately for 
downstream

I want to raise the issue of shared vs. static libraries for dpdk in vpp.

Would there be any performance penalty to going with shared?

--Tom

--
Thomas F Herbert
NFV and Fast Data Planes
Networking Group Office of the CTO
Red Hat

[vpp-dev] Query on mac learning - VLAN

[Subramanya Bhatkal Nagesh]

Hi,

We are interested in understanding if we can learn mac address on a per VLAN 
basis in VPP.

If an interface 'X' is part of VLAN 100, and interface 'Y' is part of VLAN 200
A host 'H' with mac "00:10:94:00:00:33" is behind a switch, ports of which are 
part of both VLAN 100 and VLAN 200.
Mac should be learnt for host 'H' on both VLAN's 100 and 200.

Traffic shall be forwarded to host based on the VLAN.

In the current implementation, by default the mac is learnt on the last learnt 
interface

DBGvpp# show l2fib verbose
Mac-Address BD-Idx If-Idx BSN-ISN Age(min) static filter bvi 
Interface-Name
 08:00:27:aa:bb:211  4  0/1  -   -  - - 
GigabitEthernet0/8/0.100
L2FIB total/learned entries: 1/1  Last scan time: 0.e0sec  Learn limit: 
4194304

DBGvpp# l2fib add 08:00:27:aa:bb:21 100 GigabitEthernet0/9/0.100

DBGvpp# show l2fib verbose
Mac-Address BD-Idx If-Idx BSN-ISN Age(min) static filter bvi 
Interface-Name
 08:00:27:aa:bb:211  3  0/0  -   -  - - 
GigabitEthernet0/9/0.100
L2FIB total/learned entries: 1/0  Last scan time: 0.e0sec  Learn limit: 
4194304

Expected Table:
Table shall look something like below
Mac-Address BD-Idx If-Idx BSN-ISN Age(min) static filter bvi 
Interface-Name
 00:10:94:00:00:331  4  0/0  -   -  - - 
GigabitEthernet0/8/0.100
00:10:94:00:00:331  3  0/0  -   -  - - 
GigabitEthernet0/9/0.200
L2FIB total/learned entries: 2/2  Last scan time: 0.e0sec  Learn limit: 
4194304


Regards,
Subramanya
Radisys

[vpp-dev] dpdk shared libraries

[Thomas F Herbert]

OVS as of 9.x now uses shared libraries with dpdk packaged separately 
for downstream


I want to raise the issue of shared vs. static libraries for dpdk in vpp.

Would there be any performance penalty to going with shared?

--Tom

--
*Thomas F Herbert*
NFV and Fast Data Planes
Networking Group Office of the CTO
*Red Hat*

[vpp-dev] IKEv2 VPN tunnel working in one direction

[via Lists.Fd.Io]

Hi,
I had configured VPP with IKEv2 & the other end is Strongswan.VPN tunnel is 
established successfully. When I start ping from Strongswan I can see 
encapsulated packet towards VPP & VPP is able to successfilly decap it & give 
it to host. But ICMP reply from host is by passed through VPP.When I checked 
FIB entry, there is default route to Gateway. Is there any configuration I am 
missing?Looking forward for the reply.
VPP IKev2 Configuration:
vpp# show ikev2 profileprofile pr1  auth-method shared-key-mic auth data Vpp123 
 local id-type fqdn data vpp.home  remote id-type fqdn data 
roadwarrior.vpn.example.com  local traffic-selector addr 172.16.124.0 - 
172.16.124.255 port 0 - 65535 protocol 0  remote traffic-selector addr 
172.16.4.0 - 172.16.4.255 port 0 - 65535 protocol 0
vpp# show ikev2 sa iip 147.75.91.166 ispi 6079d8558bccd119 rip 147.75.201.30 
rspi a2620372c871ac92 encr:aes-cbc-256 prf:hmac-sha1 integ:sha1-96 
dh-group:modp-1024  nonce 
i:c71dcee5b089e058ab7614dee2638711e113827ce4109e4e21a276fb4b40980d        
r:95f1a2d36da6f64122a5c887bf84f5f35f2424344a4543c3e1dd1e1e2fde8cd2  SK_d    
4d4b5affe49fcbb38c2bff3e0e9aac74c8a93320  SK_a  
i:0b3784270c25203a811b78965f5f5f2ee5fe2ea0        
r:4d035d81c68b40c5bcd8b69f35ee650a9b423aa5  SK_e  
i:1db2b5cbf1204b38834abb0b557f4df4b1e8db70cff48e96133fae5ea970ef3e        
r:8a4093c2ce61e8e26ace85ccd897c6c0a187a9e1be8cdd6f48c7e38c5af2a96f  SK_p  
i:4baa1726ffed5116952aebe9ac8a31b1eb9e5968        
r:710ab0d9d64c7ad34c8ef8e5f2919214e7ef4687  identifier (i) fqdn 
roadwarrior.vpn.example.com  identifier (r) fqdn vpp.home  child sa 0:    
encr:aes-cbc-192 integ:sha1-96 esn:no    spi(i) c6a49aef spi(r) d7541012    
SK_e  i:26b9ff196e7afd401799eeddb720f924bf185983797afa55          
r:56e12e9f1dc8cee609d8405786e4a61a1d57f5c6ebdee909    SK_a  
i:a1de061de25c82295f3c7f1f12eeb32e4048d28b          
r:e22af26875ab79698b260b832993c63bacce6a59    traffic selectors (i):      0 
type 7 protocol_id 0 addr 172.16.4.0 - 172.16.4.255 port 0 - 65535    traffic 
selectors (r):      0 type 7 protocol_id 0 addr 172.16.124.0 - 172.16.124.255 
port 0 - 65535 iip 147.75.91.166 ispi 6079d8558bccd119 rip 147.75.201.30 rspi 
a2620372c871ac92

Thanks,Saurabh

[vpp-dev] The VCL server CPU utilization is 100% ， if there is no message to the epoll_wait

[xyxue]

Hi guys，

I‘m testing the VCL . The VCL server CPU utilization is 100% ,if there is no 
message to the epoll_wait. Is there anything I can do to slove it?
Is there  the same problem in the VCL client?

Thanks,
Xyxue