Experimental Physics and
Industrial Control System

Michael Davidsaver via Tech-talk <tech-talk at aps.anl.gov> · Thu, 18 Feb 2021 21:43:18 -0800

On 2/18/21 9:04 PM, Li, Ji wrote:
> Thank you Michael. You mentioned "negotiate numeric IDs", does that mean ('ca.sid', 'ca.cid' and 'ca.ioid') are dynamically allocated for the PVs.

Yes, all of these IDs are dynamically allocated.  So you can't
rely on them being stable across re-connection.

> Is it possible to filter out some specific PVs?

Not easily, and not if you start capturing mid-stream.  That is,
if a capture doesn't include the create channel, get request,
or subscription request messages.

I don't have a general recipe for troubleshooting CA.  Can you
say something about what (mis)behavior you are trying to trap?

> The record test/ca_all_dbr.pcapng shows counting sids for the PVs in successive connections, but cids were never changed. Is this guaranteed as protocol implementation or just a single occurrence? I'll have to trace some PVs and it may run for days or even longer.

You are seeing an implementation detail of libca, which starts
allocating ID from 0 each time a context is created.  I was
running "caget" in a shell script loop, so each invocation
started a new process, which started over from 0.

> Best,
> 
> -Ji Li
> 
> ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
> *发件人:* Michael Davidsaver <mdavidsaver at gmail.com>
> *发送时间:* 2021年2月18日 23:15
> *收件人:* Li, Ji <liji at bnl.gov>
> *抄送:* Mark Rivers <rivers at cars.uchicago.edu>; tech-talk at aps.anl.gov <tech-talk at aps.anl.gov>
> *主题:* Re: 回复: How to capture messages generated by PyEpics?
>  
> On 2/18/21 7:42 PM, Li, Ji via Tech-talk wrote:
>> Yes, I used PV name (ca.pv) or the IP address of the workstation from which messages were generated (ip.addr) as the filter.
> 
> Try plain "ca" as a filter.  CA is a highly stateful protocol, which complicates analysis.
> It only includes the text PV name in the search and create channel messages.  Beginning
> with the create channel reply numeric IDs are negotiated ('ca.sid', 'ca.cid'), and then
> another 'ca.ioid' for individual get/put operations.
> 
> eg. following on Get operation all the way through might look like:
> 
>> (ca.pv=="test:ao") || (ca.cid == 1) || (ca.sid == 0) || (ca.ioid == 3)
> 
> The cashark repository contains a few pre-recorded capture files with known traffic.
> eg. test/ca_all_dbr.pcapng records my running "caget" in a loop for all valid DBR
> type codes.
> 
>> Just tried to capture in the same way on the workstation with Wireshark 2.6.20. Got the same result.
> 
> This is the version I have atm. (from Debian 10), so I'm confident it should works.
> 
> One possible complication is using a non-standard port.  The nature of CA messages
> prevents automatic classification.  If your CA traffic is on a non-standard port,
> you will have to manually tell wireshark about this.  Either interactively by finding
> at least one frame in the stream and following the context menu "Decode As...",
> or editing ca.lua to change the defaults.
> 
>> $ grep 5064 ca.lua 
>> utbl:add(5064, ca)
>> ttbl:add(5064, ca)
> 
> 
>> I noticed that in the received messages for both caget and caput, there were only commands "Version" 0x00, "Search" 0x06, "User" 0x14, "Host" 0x15 and "Create channel" 0x12, while I was expecting to see "Read" 0x03 and "Write" 0x04 separately.
> 
> Read and Write messages include 'ca.sid' and 'ca.ioid'.
> 
>> Best,
>> 
>> -Ji Li
>> 
>> ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>> *发件人:* Mark Rivers <rivers at cars.uchicago.edu>
>> *发送时间:* 2021年2月18日 22:17
>> *收件人:* Li, Ji <liji at bnl.gov>
>> *抄送:* tech-talk at aps.anl.gov <tech-talk at aps.anl.gov>
>> *主题:* Re: How to capture messages generated by PyEpics?
>>  
>> I don’t think the Wireshark version will matter. Have you put a filter on what Wireshark captures, or on what is selected to display?
>> 
>> Mark
>> 
>> 
>> Sent from my iPhone
>> 
>> On Feb 18, 2021, at 8:26 PM, Li, Ji <liji at bnl.gov> wrote:
>> 
>> 
>> Thanks Mark. I expect Wireshark to be install & play. The one I used is an old installation on an old server. Will find another one and try it out.
>> 
>> Best,
>> 
>> -Ji Li
>> 
>> ________________________________
>> 发件人: Mark Rivers <rivers at cars.uchicago.edu>
>> 发送时间: 2021年2月18日 20:11
>> 收件人: Li, Ji <liji at bnl.gov>
>> 抄送: tech-talk at aps.anl.gov <tech-talk at aps.anl.gov>
>> 主题: RE: How to capture messages generated by PyEpics?
>> 
>> 
>> Ø   When I get/set from PyEpics (tried both PV module and CA module), the program was able to successfully get the up-to-date value of the PV and set new value to it,
>> 
>> Ø  > but no messages showed up. Doesn't PyEpics communicate using CA protocol messages? Are there specific filter/configuration to be done in Wireshark to show these messages?
>> 
>> 
>> 
>> There is no difference in the CA network messages using caget and PyEpics.    They both use the same libca.so library from EPICS base.
>> 
>> 
>> 
>> You must have something configured wrong in Wireshark or something.
>> 
>> 
>> 
>> Mark
>> 
>> 
>> 
>> 
>> 
>> From: Tech-talk <tech-talk-bounces at aps.anl.gov> On Behalf Of Li, Ji via Tech-talk
>> Sent: Thursday, February 18, 2021 4:52 PM
>> To: tech-talk at aps.anl.gov
>> Subject: How to capture messages generated by PyEpics?
>> 
>> 
>> 
>> Hi,
>> 
>> 
>> 
>> I'm trying to capture CA messages targeting a specific PV on the IOC server with Wireshark (v1.12.1) and Michael Davidsaver's cashark plugin (ca.pv or ip.addr as the filter). Wireshark was able to capture messages generated by caget/caput commands ran from bash. When I get/set from PyEpics (tried both PV module and CA module), the program was able to successfully get the up-to-date value of the PV and set new value to it, but no messages showed up. Doesn't PyEpics communicate using CA protocol messages? Are there specific filter/configuration to be done in Wireshark to show these messages? Thanks.
>> 
>> 
>> 
>> 
>> 
>> -Ji Li
>> 
>> 
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System