cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
alexis_jp
Explorer
Explorer
1,611 Views
Registered: ‎09-10-2019

hw_server & Vivado spamming all my network interfaces

Jump to solution

How can I stop hw_server and vivado from sending any useless UDP packets to all my interfaces?

Firewall isn't a solution, an option has to be available to disable that unnecessary and annoying traffic.

Tags (1)
0 Kudos
1 Solution

Accepted Solutions
alexis_jp
Explorer
Explorer
1,391 Views
Registered: ‎09-10-2019

As for I should stop helping Xilinx, and Xilinx support would get better, I think you praise all the "super users" too much. Having used other FPGA companies support, Xilinx's is IMHO better. Removing all us unpaid help would just flood the system with students questions, and the proper support people would be swamped.

I come from Intel. Their support is much better than Xilinx. They didn't ignore bugs, they didn't ignore wrong documentations. All the questions I had regarding Xilinx aren't support but bug reports or missing documentation.

The parallels with say Linux and Xilinx tools are also interesting, Linux is "open source" and large parts are free, but large parts you have to pay to get support. Just like Xilinx and the other FPGA vendors.
The irony that the forums are run more like open source support and you saying I should do more open source support is not lost.

That's interesting, do you consider lack of documentations or undocumented features support?

Xilinx is all but an open source company.

Bottom line, you have found something that I have not heard of being a problem before. I can not see a "switch" to disable what you see, so its a software modification your asking for.
As such, as I have suggested, you quantify what you see, and document how you have seen it, and then talk to your FAE,

Unnecessary traffic is always a problem. Even more true when it uses all the interfaces on a server. (found that https://twitter.com/azonenberg/status/973039519455895552)

But you're right, compare to others, I try to raise such "kinda issues" to solve it or at least to have an explanation. Wasting time.

Anyway, thanks for your answers.

View solution in original post

0 Kudos
11 Replies
drjohnsmith
Teacher
Teacher
1,572 Views
Registered: ‎07-09-2009
I dont know, but

A guess, is it the web pack your using ?
if so you did sign up to it "calling home " in the T&C you signed up to . Could it be that your seeing ?

After that, your going to have to give some more details

Which version Vivado,
Which OS and version
What / how have you diagnosed its vivado ?
How often do you see it doing what exactly ?

<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
alexis_jp
Explorer
Explorer
1,490 Views
Registered: ‎09-10-2019

All that doesn't matter, Vivado is sending UDP broadcast packets, it can't be just a bug.

Vivado 2019.1, Linux Centos 7

$ sudo tcpdump -nn -i enp2s0f1
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on enp2s0f1, link-type EN10MB (Ethernet), capture size 262144 bytes
13:09:15.835993 IP 10.50.0.10.1534 > 10.50.0.255.1534: UDP, length 8
13:09:30.836101 IP 10.50.0.10.1534 > 10.50.0.255.1534: UDP, length 8

How do I know it's Vivado?

$ sudo netstat -tulpn| grep 1534
udp        0      0 0.0.0.0:1534            0.0.0.0:*                           15178/vivado 

Only Vivado is using the UDP port 1534.

When hw_server is running:

$ sudo tcpdump -nn -i enp2s0f1
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on enp2s0f1, link-type EN10MB (Ethernet), capture size 262144 bytes
13:11:00.837169 IP 10.50.0.10.1534 > 10.50.0.255.1534: UDP, length 266
13:11:07.253726 IP 10.50.0.10.38979 > 10.50.0.255.1534: UDP, length 266
13:11:15.837322 IP 10.50.0.10.1534 > 10.50.0.255.1534: UDP, length 266
13:11:22.253845 IP 10.50.0.10.38979 > 10.50.0.255.1534: UDP, length 266

hw_server as a client:

$ sudo netstat -tulpn| grep 38979
udp        0      0 0.0.0.0:38979           0.0.0.0:*                           3477/hw_server

 The WebTalk is disabled.

0 Kudos
drjohnsmith
Teacher
Teacher
1,473 Views
Registered: ‎07-09-2009
So I'm guessing its advertising itself on the network for clients to pick up,

how much is it doing what ? one packet per second or 1 packet per micro second ?


<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
0 Kudos
alexis_jp
Explorer
Explorer
1,472 Views
Registered: ‎09-10-2019

That doesn't matter and I'm not interested to waste my time trying to reverse engineer what Xilinx does.

How to disable that?

0 Kudos
drjohnsmith
Teacher
Teacher
1,463 Views
Registered: ‎07-09-2009
Ok,

Please remember we are here helping because we want to, we do not work for nor receive any payment or support from Xilinx.

I would say that as this has not been reported before, there is no switch for it,
Hence my questions about "how bad it was",

If its very serious, say swamping your network, it should be repeatably, and I'd suggest raising a support case with your FAE.


<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
0 Kudos
alexis_jp
Explorer
Explorer
1,428 Views
Registered: ‎09-10-2019

@drjohnsmithI really appreciate you give your time without begin paid to help a struggling billions dollars company. Joke apart, unfortunately, without people like you Xilinx wouldn't have any other choice than improving their support. IMHO you should rather give your time for open-source projects.

Regarding my binary question, it is a pure Xilinx's app question, something that should have been documented somewhere.

Swamping or not, isn't relevant. I don't want to have useless noise traffic through all my interfaces.

You're right, it might be faster through the FAE.

Anyway, thanks ;)

drjohnsmith
Teacher
Teacher
1,405 Views
Registered: ‎07-09-2009
Thank you for your comments,

If its of interest, I am very active in open source VHDL cores,

As for I should stop helping Xilinx, and Xilinx support would get better, I think you praise all the "super users" too much. Having used other FPGA companies support, Xilinx's is IMHO better. Removing all us unpaid help would just flood the system with students questions, and the proper support people would be swamped.

The parallels with say Linux and Xilinx tools are also interesting, Linux is "open source" and large parts are free, but large parts you have to pay to get support. Just like Xilinx and the other FPGA vendors.

The irony that the forums are run more like open source support and you saying I should do more open source support is not lost.

Bottom line, you have found something that I have not heard of being a problem before. I can not see a "switch" to disable what you see, so its a software modification your asking for.
As such, as I have suggested, you quantify what you see, and document how you have seen it, and then talk to your FAE,

Let us know how you get on,
<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
0 Kudos
alexis_jp
Explorer
Explorer
1,392 Views
Registered: ‎09-10-2019

As for I should stop helping Xilinx, and Xilinx support would get better, I think you praise all the "super users" too much. Having used other FPGA companies support, Xilinx's is IMHO better. Removing all us unpaid help would just flood the system with students questions, and the proper support people would be swamped.

I come from Intel. Their support is much better than Xilinx. They didn't ignore bugs, they didn't ignore wrong documentations. All the questions I had regarding Xilinx aren't support but bug reports or missing documentation.

The parallels with say Linux and Xilinx tools are also interesting, Linux is "open source" and large parts are free, but large parts you have to pay to get support. Just like Xilinx and the other FPGA vendors.
The irony that the forums are run more like open source support and you saying I should do more open source support is not lost.

That's interesting, do you consider lack of documentations or undocumented features support?

Xilinx is all but an open source company.

Bottom line, you have found something that I have not heard of being a problem before. I can not see a "switch" to disable what you see, so its a software modification your asking for.
As such, as I have suggested, you quantify what you see, and document how you have seen it, and then talk to your FAE,

Unnecessary traffic is always a problem. Even more true when it uses all the interfaces on a server. (found that https://twitter.com/azonenberg/status/973039519455895552)

But you're right, compare to others, I try to raise such "kinda issues" to solve it or at least to have an explanation. Wasting time.

Anyway, thanks for your answers.

View solution in original post

0 Kudos
bill_s
Visitor
Visitor
362 Views
Registered: ‎04-22-2021

 I have the same problem. It appears to be a regular message from TCF - perhaps a polling message? For me it started when I downloaded my application to the Zynq device from Vitis. I am running Vitis 2020.2. I am downloading and debugging via a USB/JTAG connection and using console IO via a USB/UART connection, so there is no value in these TCF messages in my current configuration.

I would very much like to be able to stop the unwanted network traffic on the Ethernet interface, because it's getting in the way of seeing my application working.

Any suggestions welcomed. Thanks

0 Kudos
alexis_jp
Explorer
Explorer
337 Views
Registered: ‎09-10-2019

In linux, drop all the UDP packets to port 1534:

sudo iptables -I OUTPUT -p udp --dport 1534 -j DROP
bill_s
Visitor
Visitor
281 Views
Registered: ‎04-22-2021

Thank you. I will try this.

0 Kudos