linuxptp-users — Open discussion on using the linuxptp software.

Re: [Linuxptp-users] (no subject)

[Richard C. <ric...@gm...>]

On Wed, Jul 13, 2016 at 10:25:46AM +0200, Baya Oussena wrote:
> Once again, I thank you very much for your help, you saved me a lot of
> time. Indeed our architecture is Intel 82574 (driver e1000e) so I will have
> to think of an other solution.

Oh man.  That is not one of Intel's better parts.  Oh well.

If you want to use ptp4l, you are going to have to implement multiple
clocks, each with its own configuration.  Take a look at clock.h and
clock.c.  The functional interface would allow more than one clock,
but the implementation does not.  We have:

	struct clock the_clock;
	...
	clock_create()
	{
        	struct clock *c = &the_clock;
		...
		return c;
 	}

At the very least, you will have to properly allocate the new clock.
After that, you'll have to make sure it all works.  My gut feeling is
that there will be some issues to resolve...

Good luck,
Richard

Re: [Linuxptp-users] (no subject)

[Baya O. <bay...@gm...>]

Hallo Richard,

Once again, I thank you very much for your help, you saved me a lot of
time. Indeed our architecture is Intel 82574 (driver e1000e) so I will have
to think of an other solution.

My part work of work is a small part of an existing architecture and I will
have to find a solution avoiding to make others making any change.

in fact it is a complicated architecture , we have redundancy and also
duplication.
For example we could be in a situation where two connected trains share a
same part of the tract during a journey and than get split out somewhere.
so each part of the two connected train has two engines (front and rear),
and each engine has redundancy for safety reasons .. all computers Boards
may be switched on but only on computer Board is active at a time..

well, I will go through the step you already mentioned and see how I can
gain any advance from it.

Once again thank, thank very much,

Best regards,
Baya


2016-07-13 8:12 GMT+02:00 Richard Cochran <ric...@gm...>:

> On Tue, Jul 12, 2016 at 11:35:18PM +0200, Richard Cochran wrote:
> > - Another idea is to run Layer-2 and isolate using VLANs.
>
> Also:
>
> - You can try running UDP over two different VLAN interfaces.
>
> - VLAN will only work if your HW correctly identifies PTP frames with
>   a VLAN header.
>
> - Regarding the HW: Some MAC HW can only time stamp one transmit frame
>   at a time, like the Intel 82574 (driver e1000e) for example.  If you
>   have such HW, then running two ptp4l instances will *not* work.
>
> HTH,
> Richard
>

Re: [Linuxptp-users] (no subject)

[Richard C. <ric...@gm...>]

On Tue, Jul 12, 2016 at 11:35:18PM +0200, Richard Cochran wrote:
> - Another idea is to run Layer-2 and isolate using VLANs.

Also:

- You can try running UDP over two different VLAN interfaces.

- VLAN will only work if your HW correctly identifies PTP frames with
  a VLAN header.

- Regarding the HW: Some MAC HW can only time stamp one transmit frame
  at a time, like the Intel 82574 (driver e1000e) for example.  If you
  have such HW, then running two ptp4l instances will *not* work.

HTH,
Richard

Re: [Linuxptp-users] (no subject)

[Richard C. <ric...@gm...>]

On Tue, Jul 12, 2016 at 09:04:25PM +0200, Baya Oussena wrote:
> so my idea is to have two ptpd daemon as slave on the I/O board
> controller . One of the slave daemon will communicate with the front master
> and the other will communicate  with the other master at  back of the train.

Okay, so this is starting to make sense.  Another question:  Why does
the controller need both master times?  Is it for redundancy, or are
there two control loops, each synchronized to a different engine?
 
> I read about configuring domain. How could I use this idea.

So you only have one physical Ethernet port and one PHC, right?

You have a number of options, but I haven't tried any of them!

Here are a few ideas:

- There are different ideas out there about redundancy for PTP.
  Supporting any of these properly (within one process) would be quite
  some work, I should think.

- If you have only one PHC, then you must run one or both ptp4l
  instance with option "free_running 1".  You can still run your
  servos, using the data from the TIME_STATUS_NP management query.

- Simply using two different domains *might* work, if not with UDP
  then at least Layer-2 should work.

- You can also consider using two different transports to isolate the
  pairs, like UDP and Layer-2.

- Another idea is to run Layer-2 and isolate using VLANs.

Good luck,
Richard

Re: [Linuxptp-users] (no subject)

[Baya O. <bay...@gm...>]

Hallo Richard,

Thank you for your answer ... As a matter of fact,  when my colleague has
ask me that i also said that it does not make sens :-) and we would need
separate port and PHC :-) ..

I clear is our situation :

the purpose will be used on embedded software on train .

the train  has two railway engine, one on the front and the other on the
back. these two engines run a ptpd daemon as a master. both  engine
communicate independently with an I/O board controller which would run a
ptpd  daemon as a slave.


so my idea is to have two ptpd daemon as slave on the I/O board
controller . One of the slave daemon will communicate with the front master
and the other will communicate  with the other master at  back of the train.

I read about configuring domain. How could I use this idea.

Once again thank you for all,
Baya






2016-07-12 20:39 GMT+02:00 Richard Cochran <ric...@gm...>:

> Well it is possible, provided that each ptp4l instance has its own
> port, and that each port has its own PHC.
>
> But to be quite honest, your proposed use case does not make any
> sense.  What are you trying to acheive?
>
> Thanks,
> Richard
>

Re: [Linuxptp-users] (no subject)

[Richard C. <ric...@gm...>]

Well it is possible, provided that each ptp4l instance has its own
port, and that each port has its own PHC. 

But to be quite honest, your proposed use case does not make any
sense.  What are you trying to acheive?

Thanks,
Richard

Re: [Linuxptp-users] Planning release 1.7

[W.F. Y. <wen...@mo...>]

Richard，
Here i got a minor suspect issue:
Role: slave node
Delay Mechanism: E2E
Scenario: slave sent a delay_req to master,   and duration period of
issuing delay_req for next time is determined by set_tmo_random(),  it is a
random value.  If the slave node send the delay_req again prior to
receiving the first delay_resp from master, the first delay_resp will be
discarded due to having previous sequence number

Anyway, it does not bring big impact.

*Best Regards*

*Yang Wenfeng*

On Tue, Jul 12, 2016 at 7:30 PM, Richard Cochran <ric...@gm...>
wrote:

> Dear linuxptp users and developers,
>
> I am planning to release version 1.7 in one week, without any major
> new features, in order to fix the regressions in version 1.6.  The two
> bugs that need fixing are:
>
> - time stamping in one step P2P mode
> - bogus announce timer on the UDS interface
>
> If you know of any other problems with v1.6 or in the current git
> head, please let me know right away.
>
> Thanks,
> Richard
>
>
> ------------------------------------------------------------------------------
> What NetFlow Analyzer can do for you? Monitors network bandwidth and
> traffic
> patterns at an interface-level. Reveals which users, apps, and protocols
> are
> consuming the most bandwidth. Provides multi-vendor support for NetFlow,
> J-Flow, sFlow and other flows. Make informed decisions using capacity
> planning
> reports.http://sdm.link/zohodev2dev
> _______________________________________________
> Linuxptp-users mailing list
> Lin...@li...
> https://lists.sourceforge.net/lists/listinfo/linuxptp-users
>

[Linuxptp-users] (no subject)

[Baya O. <bay...@gm...>]

Hallo Dear linuxptp-users,

Please could any one from you help me answer this question.

For our Project Needs,  I would like to have the possibility of doing the
following :

I would like to run two demons of ptp4l on one machine as a slave , that is
slave 1 and slave 2

and two demons of ptp4l on the second machine with two master:  master1
,master 2.

I would like to have : slave1 synchronising with master1 and
                                  slave2 synchronising with master2.


Is this that possible?

Thank you in advance for your help,

Baya

Re: [Linuxptp-users] Planning release 1.7

[Richard C. <ric...@gm...>]

On Tue, Jul 12, 2016 at 10:36:29PM +0800, W.F. YANG wrote:
> Richard，
> Here i got a minor suspect issue:
> Role: slave node
> Delay Mechanism: E2E
> Scenario: slave sent a delay_req to master,   and duration period of
> issuing delay_req for next time is determined by set_tmo_random(),  it is a
> random value.  If the slave node send the delay_req again prior to
> receiving the first delay_resp from master, the first delay_resp will be
> discarded due to having previous sequence number

This is the expected behavior.  The timing of the delay request
messages does not depend on the reponse from the master.

(Consider what would happen if the master never responds...)

Thanks,
Richard

Re: [Linuxptp-users] [Linuxptp-devel] Planning release 1.7

[Richard C. <ric...@gm...>]

On Tue, Jul 12, 2016 at 12:14:47PM +0000, Jesuiter, Henry (ALC NetworX GmbH) wrote:
> Hello,
> 
> as a bug fix release, it may be useful if you consider my patch from last Friday ([PATCH] Fix data type for return value of vasprintf()), 
> regarding the the data type issue in util.c. 
> 
> 'len2' is of type 'size_t' in util.c:451, but vasprintf in line 457 returns '-1' on error which is checked in line 460. Currently 
> this check will always fail (regardless of the return value of vasprintf()), according to its declaration as unsigned int.

Yes, that is a bug.  I will include the fix in 1.7.

Thanks,
Richard

[Linuxptp-users] Planning release 1.7

[Richard C. <ric...@gm...>]

Dear linuxptp users and developers,

I am planning to release version 1.7 in one week, without any major
new features, in order to fix the regressions in version 1.6.  The two
bugs that need fixing are:

- time stamping in one step P2P mode
- bogus announce timer on the UDS interface

If you know of any other problems with v1.6 or in the current git
head, please let me know right away.

Thanks,
Richard

Re: [Linuxptp-users] Does ptp4l support unicast ?

[Richard C. <ric...@gm...>]

On Sun, Jul 10, 2016 at 12:24:28AM +0800, W.F. YANG wrote:
> Does the ptp4l support ip unicast instead of ip multicast for all
> transmission, not just for delay request ?

No.

> Besides, is it a portion of ptp protocol standard ?

Yes, unicast is an option within the IEEE 1588 standard.


HTH,
Richard

[Linuxptp-users] Does ptp4l support unicast ?

[W.F. Y. <wen...@mo...>]

Hello,
Does the ptp4l support ip unicast instead of ip multicast for all
transmission, not just for delay request ?
Besides, is it a portion of ptp protocol standard ?

*Best Regards*

*Yang Wenfeng*

Re: [Linuxptp-users] Problems with ptp4l via a Cisco Nexus 5000 Switch

[Richard C. <ric...@gm...>]

BTW, in the log I noticed a bug (not related to your problem):

On Fri, Jul 01, 2016 at 03:26:05PM +0200, Christian wrote:
> ptp4l[622912.203]: port 0: INITIALIZING to LISTENING on INITIALIZE
> ptp4l[622912.209]: port 0: setting asCapable
> ptp4l[622918.492]: port 0: announce timeout
> ptp4l[622925.985]: port 0: announce timeout
> ptp4l[622932.071]: port 0: announce timeout

Port 0 is the UDS port and should never see an announce message
timeout.  This is a regression in v1.6.  I just posted a fix for it on
the linuxptp-devel mailing list.

Thanks,
Richard

Re: [Linuxptp-users] Problems with ptp4l via a Cisco Nexus 5000 Switch

[Richard C. <ric...@gm...>]

On Fri, Jul 01, 2016 at 03:26:05PM +0200, Christian wrote:
> I have attached 3 .txt files with the logs. Again, 96 is the master server
> and 94 and 92 are the slaves.
> And I have attached a pcap file of the Server 94(not captured at the same
> moment, but the behavior was the same).

Sorry about the delay.  I finally took a look at your problem...

In the PCAP, the Announce messages are only sent by the Cisco every
seven seconds.  The PCAP frames are truncated, so we can't see what
the Cisco advertises for the logMessageInterval (logAnnounceInterval).
I guess it is set to 3 (2^3 = 8 seconds).

Anyhow, you can fix this by configuring the Cisco and your client to
the same logAnnounceInterval value.

Thanks,
Richard

Re: [Linuxptp-users] Problems with ptp4l via a Cisco Nexus 5000 Switch

[Christian <chr...@fr...>]

I have attached 3 .txt files with the logs. Again, 96 is the master 
server and 94 and 92 are the slaves.
And I have attached a pcap file of the Server 94(not captured at the 
same moment, but the behavior was the same).



Am 30.06.2016 um 21:05 schrieb Richard Cochran:
> On Thu, Jun 30, 2016 at 11:32:49AM -0400, Dale Smith wrote:
>> I think it would be far better to submit the logs as text instead of
>> images.  Those images are very difficult to read and impossible to edit
>> (the text).
> +1
>
> In addition, can you post a short pcap file showing the PTP traffic on
> the non-working port?
>
> Thanks,
> Richard

Re: [Linuxptp-users] Problems with ptp4l via a Cisco Nexus 5000 Switch

[Richard C. <ric...@gm...>]

On Thu, Jun 30, 2016 at 11:32:49AM -0400, Dale Smith wrote:
> I think it would be far better to submit the logs as text instead of
> images.  Those images are very difficult to read and impossible to edit
> (the text).

+1

In addition, can you post a short pcap file showing the PTP traffic on
the non-working port?

Thanks,
Richard

Re: [Linuxptp-users] Problems with ptp4l via a Cisco Nexus 5000 Switch

[Dale S. <dal...@gm...>]

On Thu, Jun 30, 2016 at 8:14 AM, Christian <
chr...@fr...> wrote:

> Here are pictures of the logs:
> Server s0002796 is the master s0002792 ands0002794 are the slaves.
>
> ptplog7start is the log it produced right after starting ptp4l.:
> https://s32.postimg.org/i6af9ejx1/ptplog7start.png
> ptplog7 is shortly thereafter:
> https://s32.postimg.org/bnsuv6rat/ptplog7.png
> ptplog7_2 is a bit later.: https://s32.postimg.org/jmtjfahhh/ptplog7_2.png
>

Howdy Christian,

I think it would be far better to submit the logs as text instead of
images.  Those images are very difficult to read and impossible to edit
(the text).

-Dale

Re: [Linuxptp-users] Problems with ptp4l via a Cisco Nexus 5000 Switch

[Christian <chr...@fr...>]

Hello,

As I have not made any progress in the matter, I have made a new log, so 
you can maybe see better where the Problem is.
I have taken out the path trace enable option and upped the logging 
level to 7, else all the configuration is still the same.


Here are pictures of the logs:
Server s0002796 is the master s0002792 ands0002794 are the slaves.

ptplog7start is the log it produced right after starting ptp4l.: 
https://s32.postimg.org/i6af9ejx1/ptplog7start.png
ptplog7 is shortly thereafter: https://s32.postimg.org/bnsuv6rat/ptplog7.png
ptplog7_2 is a bit later.: https://s32.postimg.org/jmtjfahhh/ptplog7_2.png


After a while I checked the PHC clocks via phc_ctl and 94 seemed to be 
synchronising, while 92 was not.


Thanks again for every little bit of help.

Greetings,
Christian

Am 23.06.2016 um 14:08 schrieb Christian:
> Hello,
>
> I'm currently trying to set up a ptp4l session between 3 servers, which are connected via a Cisco Nexus 5000 switch
>
>
> Server A is supposed to be the grand master, Server B and C should be the slaves.
>
>
>
> The Grandmaster is working fine, the Switch does accept it, but the Slaves are not working properly.
>
> Here is the ptp log of one of the slave servers:
>
> @s0002794:~$ ptp4l[618557.966]: selected /dev/ptp0 as PTP clock
> ptp4l[618557.967]: driver changed our HWTSTAMP options
> ptp4l[618557.967]: tx_type   1 not 1
> ptp4l[618557.967]: rx_filter 1 not 12
> ptp4l[618557.967]: port 1: INITIALIZING to LISTENING on INITIALIZE
> ptp4l[618557.967]: port 0: INITIALIZING to LISTENING on INITIALIZE
> ptp4l[618564.664]: driver changed our HWTSTAMP options
> ptp4l[618564.664]: tx_type   1 not 1
> ptp4l[618564.664]: rx_filter 1 not 12
> ptp4l[618564.664]: selected best master clock a0369f.fffe.a1b68c		<-- MAC Addr. of the Slave server
> ptp4l[618564.840]: port 1: new foreign master 002a6a.fffe.ac97fc-16		<-- MAC Addr. of the Switch port
> ptp4l[618571.137]: driver changed our HWTSTAMP options
> ptp4l[618571.137]: tx_type   1 not 1
> ptp4l[618571.137]: rx_filter 1 not 12
> ptp4l[618571.137]: selected best master clock a0369f.fffe.a1b68c
> ptp4l[618578.192]: driver changed our HWTSTAMP options
> ptp4l[618578.192]: tx_type   1 not 1
> ptp4l[618578.192]: rx_filter 1 not 12
> ptp4l[618578.192]: selected best master clock a0369f.fffe.a1b68c
> ptp4l[618580.850]: selected best master clock a0369f.fffe.a1b688		<-- MAC Addr. of the Master server
> ptp4l[618580.850]: foreign master not using PTP timescale
> ptp4l[618580.850]: running in a temporal vortex
> ptp4l[618580.850]: port 1: LISTENING to UNCALIBRATED on RS_SLAVE
> ptp4l[618582.669]: port 1: minimum delay request interval 2^4
> ptp4l[618582.851]: master offset    1299422 s0 freq   +4989 path delay    -11798
> ptp4l[618584.852]: master offset    1321175 s1 freq  +15865 path delay     -9425
> ptp4l[618587.572]: driver changed our HWTSTAMP options
> ptp4l[618587.572]: tx_type   1 not 1
> ptp4l[618587.572]: rx_filter 1 not 12
> ptp4l[618587.572]: port 1: UNCALIBRATED to LISTENING on ANNOUNCE_RECEIPT_TIMEOUT_EXPIRES
> ptp4l[618587.572]: selected best master clock a0369f.fffe.a1b68c
> ptp4l[618594.597]: driver changed our HWTSTAMP options
> ptp4l[618594.597]: tx_type   1 not 1
> ptp4l[618594.597]: rx_filter 1 not 12
> ptp4l[618594.597]: selected best master clock a0369f.fffe.a1b68c
> ptp4l[618596.851]: selected best master clock a0369f.fffe.a1b688
> ptp4l[618596.851]: foreign master not using PTP timescale
> ptp4l[618596.851]: running in a temporal vortex
> ptp4l[618596.851]: port 1: LISTENING to UNCALIBRATED on RS_SLAVE
> ptp4l[618598.850]: master offset      43778 s2 freq  +37754 path delay         0
> ptp4l[618598.850]: port 1: UNCALIBRATED to SLAVE on MASTER_CLOCK_SELECTED
> ptp4l[618600.852]: master offset       6710 s2 freq  +25787 path delay         0
>
> there also was this line every one in a while:
>
> ptp4l[5828.626]: port 1: minimum delay request interval 2^^4
>
>
> I am starting the Master with a config File.
>   ptp4l -f /etc/ptp.config
>
> with following config file:
>
> [global]
> verbose               1
> path_trace_enabled    1
> time_stamping         hardware
> priority1             1
> priority2             1
> [eth4]
>
>
> The Slaves are started via: ptp4l -A -m -i eth4 -s
>
> I also wrote a config file for them, but if I use the file, they do select themself as the best clock.
>
>
>
> rupp@s0002794:~$ ptp4l[619043.020]: selected best master clock a0369f.fffe.a1b68c <--MAC Addr. of the Slave
> ptp4l[619050.271]: selected best master clock a0369f.fffe.a1b68c
> ptp4l[619056.348]: selected best master clock a0369f.fffe.a1b68c
> ptp4l[619062.726]: selected best master clock a0369f.fffe.a1b68c
>
> with following config File:
>
> [global]
> verbose               1
> path_trace_enabled    1
> time_stamping         hardware
> slaveOnly             1
> priority1             255
> priority2             255
> [eth4]
>
>
> Phc2sys is running on both, the Master and the Slaves.
> Master: phc2sys -s CLOCK_REALTIME -c eth4 -w &
> Slaves: phc2sys -s eth4 -w &
>
>
> Here is the configuration of the Switch:
>
>
> ptp brief:
>
> PTP port status
> -----------------------
> Port         State
> -------  --------------
> Eth1/17     Master 	<--Server C
> Eth1/19     Master	<--Server B
> Eth1/31     Slave	<--Server A
>
> ptp clock:
>
> PTP Device Type: Boundary clock
> Clock Identity :  00:2a:6a:ff:fe:ac:97:fc
> Clock Domain: 0
> Number of PTP ports: 3
> Priority1 : 2
> Priority2 : 2
> Clock Quality:
>          Class : 248
>          Accuracy : 254
>          Offset (log variance) : 65535
> Offset From Master : 467
> Mean Path Delay : 33580
> Steps removed : 1
> Local clock time:Thu Jun 23 13:52:14 2016
>
> Slave port:
> PTP Port Dataset: Eth1/17
> Port identity: clock identity:  00:2a:6a:ff:fe:ac:97:fc
> Port identity: port number: 16
> PTP version: 2
> Port state: Master
> VLAN info:  1
> Delay request interval(log mean): 4
> Announce receipt time out: 2
> Peer mean path delay: 0
> Announce interval(log mean): 3
> Sync interval(log mean): 1
> Delay Mechanism: End to End
>
> Master port:
> PTP Port Dataset: Eth1/31
> Port identity: clock identity:  00:2a:6a:ff:fe:ac:97:fc
> Port identity: port number: 30
> PTP version: 2
> Port state: Slave
> VLAN info:  1
> Delay request interval(log mean): 4
> Announce receipt time out: 2
> Peer mean path delay: 0
> Announce interval(log mean): 3
> Sync interval(log mean): 1
> Delay Mechanism: End to End
>
>
> Thank you in advance.
>
>
> Greeting,
> Christian
>
>
> ------------------------------------------------------------------------------
> Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
> Francisco, CA to explore cutting-edge tech and listen to tech luminaries
> present their vision of the future. This family event has something for
> everyone, including kids. Get more information and register today.
> http://sdm.link/attshape
>
>
> _______________________________________________
> Linuxptp-users mailing list
> Lin...@li...
> https://lists.sourceforge.net/lists/listinfo/linuxptp-users

Re: [Linuxptp-users] Master Clock Choice

[Vanderpool, C. <cly...@ax...>]

Just realized that the ping statistics I posted are reversed.  The correct
info is...

*2-A-->2-G-->Ground Gateway*
Average ping time - 1.203 micro sec
Stand Dev - .342 micro sec

*3-A-->3-G-->Ground Gateway*
Average ping time - 67.806 micro sec
Stand Dev - 6.333 micro sec

Sorry for the confusion.

Clyde

On Thu, Jun 23, 2016 at 11:57 AM, Vanderpool, Clyde <
cly...@ax...> wrote:

> PTP users,
>
> I have a question about how the choice of a parent clock is decided.  I
> have included a block diagram of the LAN that I have synchronized with
> ptp4l (software time stamping) in the attachment of this email.  Names of
> various 'links' are inside the green squares.
>
> The left side of the diagram is synchronized first, with the Ground
> Gateway assuming the Grand Master role.  Then the right side of the diagram
> is brought up.  On the right side link 1-A (which has no direct connection
> to the left side) and the Air Gateway will always pick link 3-A as it's
> parent boundary clock (despite the order in which they are synced to the
> LAN).  I find this strange due to the fact that the path to the Grand
> Master needs to go over a radio (3-A-->3-G-->Ground Gateway) instead of
> going through a direct Ethernet connection (2-A-->2-G-->Ground Gateway).
>
> I then tried to take link 3-G offline.  This caused links 1-A, 3-A, and
> the Air Gateway to pick 2-A as their parent boundary clock.  This makes
> sense, as 2-A offers the only connection to the Grand Master.  When I bring
> back link 3-G, link 3-A picks it as it's parent boundary clock.  This makes
> sense to me as it requires less steps to get to the Grand Master.  What
> does not make sense to me is that both link 1-A and the Air Gateway drop
> link 2-A as their parent and once again sync to link 3-A despite the fact
> that the steps removed from the Grand Master remains the same.
>
> I performed a ping test to try see how long it takes to send/receive
> information on both of the previous pathways to the Grand Master.  The
> results are as follows.
>
> *3-A-->3-G-->Ground Gateway *
>
> Average ping time - 1.203 micro sec
> Stand Dev - .342 micro sec
>
> *2-A-->2-G-->Ground Gateway*
>
> Average ping time - 67.806 micro sec
> Stand Dev - 6.333 micro sec
>
>
> I was just wondering why the two machines with no direct connection to the
> Grand Master would consistently choose the longer pathway (time wise) to
> the Grand Master.  Is there something about the Best Master Clock Algorithm
> that makes this so?
>
> I apologize if this is a bit wordy and hard to follow, but I cannot reason
> out how the choice of parent clocks is made.  Any information is greatly
> appreciated and if anyone needs any further clarification please feel free
> to ask.  Thank you in advance!
>
> Clyde
>
>
>
>

[Linuxptp-users] Master Clock Choice

[Vanderpool, C. <cly...@ax...>]

PTP users,

I have a question about how the choice of a parent clock is decided.  I
have included a block diagram of the LAN that I have synchronized with
ptp4l (software time stamping) in the attachment of this email.  Names of
various 'links' are inside the green squares.

The left side of the diagram is synchronized first, with the Ground Gateway
assuming the Grand Master role.  Then the right side of the diagram is
brought up.  On the right side link 1-A (which has no direct connection to
the left side) and the Air Gateway will always pick link 3-A as it's parent
boundary clock (despite the order in which they are synced to the LAN).  I
find this strange due to the fact that the path to the Grand Master needs
to go over a radio (3-A-->3-G-->Ground Gateway) instead of going through a
direct Ethernet connection (2-A-->2-G-->Ground Gateway).

I then tried to take link 3-G offline.  This caused links 1-A, 3-A, and the
Air Gateway to pick 2-A as their parent boundary clock.  This makes sense,
as 2-A offers the only connection to the Grand Master.  When I bring back
link 3-G, link 3-A picks it as it's parent boundary clock.  This makes
sense to me as it requires less steps to get to the Grand Master.  What
does not make sense to me is that both link 1-A and the Air Gateway drop
link 2-A as their parent and once again sync to link 3-A despite the fact
that the steps removed from the Grand Master remains the same.

I performed a ping test to try see how long it takes to send/receive
information on both of the previous pathways to the Grand Master.  The
results are as follows.

*3-A-->3-G-->Ground Gateway *

Average ping time - 1.203 micro sec
Stand Dev - .342 micro sec

*2-A-->2-G-->Ground Gateway*

Average ping time - 67.806 micro sec
Stand Dev - 6.333 micro sec


I was just wondering why the two machines with no direct connection to the
Grand Master would consistently choose the longer pathway (time wise) to
the Grand Master.  Is there something about the Best Master Clock Algorithm
that makes this so?

I apologize if this is a bit wordy and hard to follow, but I cannot reason
out how the choice of parent clocks is made.  Any information is greatly
appreciated and if anyone needs any further clarification please feel free
to ask.  Thank you in advance!

Clyde

Re: [Linuxptp-users] Problems with ptp4l via a Cisco Nexus 5000 Switch

[Christian <chr...@fr...>]

I took it out and it still has the same Problem.

Am 23.06.2016 um 15:24 schrieb Richard Cochran:
> You have path_trace_enabled, but why?  That is only for use in gPTP
> according to IEEE 802.1AS-2011.
>
>> >Here is the configuration of the Switch:
>> >PTP Device Type: Boundary clock
> Your switch is a BC.  Maybe it cannot deal with the path trace
> option.

Re: [Linuxptp-users] Frequency Adjustment

[Vanderpool, C. <cly...@ax...>]

By far the best explanation I have seen.  Thank you!

On Thu, Jun 23, 2016 at 9:13 AM, Richard Cochran <ric...@gm...>
wrote:

> On Wed, Jun 22, 2016 at 08:16:20AM -0400, Vanderpool, Clyde wrote:
> > I guess a little bit of both.  From looking things up on google or
> > different forums all I could find was kind of circular (i.e. 'frequency
> > adjustment is the computer adjusting it's frequency')  I figured it had
> > something to do with the clock pulse of the individual machines but I
> don't
> > want to guess.  I have to present some information and I want to make
> sure
> > I sound reasonably intelligent.  Any info would be appreciated.
>
> In general, there are two important variables when synchronizing two
> clocks.  The "offset" is the instantaneous time difference between the
> two clocks.  If you perfectly synchronize two wrist watches by pushing
> their pins in at exactly the same, then at that moment, the offset
> will be zero.  After that, the two watches will slowly drift away from
> each other.
>
> This happens because the quartz crystals oscillate at slightly
> different rates.  Another word for the rate is "frequency".  The
> difference between the rates of two clocks can be represented as a
> fraction or percentage.  Normally the difference is small, and so you
> will see "parts per million" (PPM) and "parts per billion" (ppb) used
> instead of percentage.
>
> Regarding the ptp4l program, it is not enough for it to simply correct
> the time offset, it also must change the rate of the local clock to
> match that of the master clock.  The program uses the PTP to estimate
> the time and frequency offsets from the master and correct them by
> applying a frequency adjustment to the local clock.  The amount of
> frequency adjustment is shown in the log in units of ppb.
>
> There are tons of papers to read about this topic.  This page and the
> site might interest you:
>
>   http://www.ntp.org/ntpfaq/NTP-s-sw-clocks-quality.htm
>
> Or try this classic David Mills paper:
>
>   https://www.eecis.udel.edu/~mills/database/papers/time.pdf
>
> Here are all of the Mills papers:
>
>   https://www.eecis.udel.edu/~mills/papers.html
>
> These two were helpful to me:
>
> @TechReport{mills1993precis,
>   author =       {David Mills},
>   title =        {Precision synchronization of computer network clocks},
>   institution =  {University of Delaware Electrical Engineering
> Department},
>   year =         1993,
>   month =        {November}}
>
> @TechReport{mills1992model,
>   author =       {David Mills},
>   title =        {Modelling and analysis of computer network clocks},
>   institution =  {University of Delaware Electrical Engineering
> Department},
>   year =         1992,
>   month =        {May}}
>
> HTH,
> Richard
>

Re: [Linuxptp-users] Problems with ptp4l via a Cisco Nexus 5000 Switch

[Richard C. <ric...@gm...>]

On Thu, Jun 23, 2016 at 02:08:02PM +0200, Christian wrote:
> there also was this line every one in a while:
> 
> ptp4l[5828.626]: port 1: minimum delay request interval 2^^4

That is informational only, not an error.

> I am starting the Master with a config File.
>  ptp4l -f /etc/ptp.config
> 
> with following config file:
> 
> [global]
> verbose               1
> path_trace_enabled    1

...

> The Slaves are started via: ptp4l -A -m -i eth4 -s
> with following config File:
> 
> [global]
> verbose               1
> path_trace_enabled    1

You have path_trace_enabled, but why?  That is only for use in gPTP
according to IEEE 802.1AS-2011.

> Here is the configuration of the Switch:
> PTP Device Type: Boundary clock

Your switch is a BC.  Maybe it cannot deal with the path trace
option.

Thanks,
Richard

Re: [Linuxptp-users] Frequency Adjustment

[Richard C. <ric...@gm...>]

On Wed, Jun 22, 2016 at 08:16:20AM -0400, Vanderpool, Clyde wrote:
> I guess a little bit of both.  From looking things up on google or
> different forums all I could find was kind of circular (i.e. 'frequency
> adjustment is the computer adjusting it's frequency')  I figured it had
> something to do with the clock pulse of the individual machines but I don't
> want to guess.  I have to present some information and I want to make sure
> I sound reasonably intelligent.  Any info would be appreciated.

In general, there are two important variables when synchronizing two
clocks.  The "offset" is the instantaneous time difference between the
two clocks.  If you perfectly synchronize two wrist watches by pushing
their pins in at exactly the same, then at that moment, the offset
will be zero.  After that, the two watches will slowly drift away from
each other.

This happens because the quartz crystals oscillate at slightly
different rates.  Another word for the rate is "frequency".  The
difference between the rates of two clocks can be represented as a
fraction or percentage.  Normally the difference is small, and so you
will see "parts per million" (PPM) and "parts per billion" (ppb) used
instead of percentage.

Regarding the ptp4l program, it is not enough for it to simply correct
the time offset, it also must change the rate of the local clock to
match that of the master clock.  The program uses the PTP to estimate
the time and frequency offsets from the master and correct them by
applying a frequency adjustment to the local clock.  The amount of
frequency adjustment is shown in the log in units of ppb.

There are tons of papers to read about this topic.  This page and the
site might interest you:

  http://www.ntp.org/ntpfaq/NTP-s-sw-clocks-quality.htm

Or try this classic David Mills paper:

  https://www.eecis.udel.edu/~mills/database/papers/time.pdf

Here are all of the Mills papers:

  https://www.eecis.udel.edu/~mills/papers.html

These two were helpful to me:

@TechReport{mills1993precis,
  author = 	 {David Mills},
  title = 	 {Precision synchronization of computer network clocks},
  institution =  {University of Delaware Electrical Engineering Department},
  year = 	 1993,
  month = 	 {November}}

@TechReport{mills1992model,
  author = 	 {David Mills},
  title = 	 {Modelling and analysis of computer network clocks},
  institution =  {University of Delaware Electrical Engineering Department},
  year = 	 1992,
  month = 	 {May}}

HTH,
Richard