[ntp:questions] Re: which 2.6 linux PPS API for oncore modules??

[Richard B. Gilbert]

mike wrote:

> Richard B. Gilbert wrote:
>
>> mike wrote:
>>
>>> I have just got an ONCORE VP and am trying to tie it
>>> to my Mandrake 2.6.8.1 distrib with ntp 4.2.0.
>>> I have reconfigured, with ./configure --enable-ONCORE  and connected 
>>> it to my homegrown pc's serial port 1 with the modules PPS output is 
>>> cabled to the DCD line.
>>>
>>> I have the links:
>>> [root at quark dev]# ls -l oncore*
>>> lrwxrwxrwx  1 root root 5 Aug  8 15:11 oncore.pps.1 -> ttyS1
>>> lrwxrwxrwx  1 root root 5 Aug  8 15:10 oncore.serial.1 -> ttyS1
>>>
>>> and ntp.conf lines:-
>>>  server 127.127.30.1
>>>  fudge  127.127.30.1 stratum 0 refid VP-1
>>>
>>> my ntp.oncore.1 config file has
>>> [root at quark dev]# cat /etc/ntp.oncore.1
>>> #
>>> # Oncore VP configuration file
>>> #
>>> #  Long/Lat/Height are for my office window :- E 002° 16.3015 / N 
>>> 48° 47.3281 / 212.85m
>>> #
>>> # ----- mandatory lines ----------------
>>> MODE     4        <====  I tried other modes.
>>> LONG     002 16.3015
>>> LAT      48 47.3281
>>> HT       M 212.85
>>> # ----- optional lines ------------------
>>> DELAY    32 NS
>>> ASSERT
>>> MASK 0
>>>
>>> On ntpd startup I get a know the error sequence.
>>>
>>> Aug  8 09:41:23 quark ntpd[10441]: refclock_newpeer: clock type 30 
>>> invalid
>>> Aug  8 09:41:23 quark ntpd[10441]: configuration of 127.127.30.1 failed
>>>
>>> Almighty Goog has informed me that this is due to that fact that I 
>>> do not have the PPS API available, and indeed, the NTP 4.2.0 driver 
>>> source confirms.
>>>
>>> I have seen here and elsewhere that a PPSkit is not available for 
>>> 2.6, but in the source tree I see that Ulrich has created a PPS-light..
>>>  PPSkit-light-PPSAPI-alpha-1610m-2.6.5.diff
>>>
>>> That looks promising, but is in the form of an autogenerated diff patch
>>> and does not fit to my code base.
>>> ex.
>>> My 2.6.8.1 source tree linux/CREDITS has  Ulrich's contrib info at 
>>> line 3458, but the diff has that at
>>> @@ -3422,7 +3422,7 @@
>>> So if even the CREDITS aren't compatible, I hesitate to proceed.
>>> Has anyone installed this against 2.6.8.1 or later?
>>> Current Linux is 2.6.12.4, so I can go there is anyone has new 
>>> diffs, but I am reluctant to go back to 2.6.5 from which the above 
>>> patch diffs were taken.
>>>
>>> There is another linux PPS API that Rodolfo Giometti has created
>>> (http://www.enneenne.com/projects/linuxpps) which may be useful, but 
>>> I cannot see if it can be used by the 4.2.0 oncore driver without 
>>> changes to the driver.
>>> Has anyone used this succesfully with the oncore driver?
>>>
>>> Any help appreciated.
>>> Mike
>>
>>
>>
>> Grab a copy of the refclock_oncore.c  source from the latest tarball 
>> <ftp://ftp.udel.edu/pub/ntp/ntp4/snapshots/ntp-dev> and replace what 
>> you have with it.  The 4.2.0 oncore driver had some code that 
>> prevented it from building.  You are getting the "clock type 30" 
>> invalid because the oncore driver did not build.  Note that you 
>> should replace ONLY the oncore driver unless you REALLY want to run 
>> the "bleeding edge" code.
>>
>> The latest version should have the patch(es).  I ran into the same 
>> problem about nineteen months ago.  Getting it going was a nightmare 
>> and I never could have done it without extensive help from the 
>> maintainer.
>
>
> Looks like I am still missing something:
>
> I could not get to the ftp  site, so I pulled ntp-dev-4.2.0b-20050808 
> from the http site and copied its refclock_oncore.c to the ntpd dir.
> [mike at quark ntp-4.2.0]$ ls -l ntpd/*oncore.c
> -rw-r--r--  1 mike home 108678 Aug  9 15:27 ntpd/refclock_oncore.c
> [mike at quark ntp-4.2.0]$ sum ntpd/*oncore.c
> 19798   107
>
> then remake and installed the daemon but the .configure output
> still shows
>
> checking NMEA GPS receiver... yes
> checking for ONCORE Motorola VP/UT Oncore GPS... no  ???
> checking for Palisade clock... yes
>
> The driver compiles, but non of the code is included. I just see 1 int
> refclock_oncore_bs which is probably due to CLOCK_ONCORE not being 
> defined.
>
> I tried forcing a compile by defining CLOCK_ONCORE, but I get loads of 
> errors, starting with
> refclock_oncore.c:203: error: syntax error before "pps_handle_t"
> refclock_oncore.c:203: warning: no semicolon at end of struct or union
> refclock_oncore.c:204: warning: type defaults to `int' in declaration 
> of `pps_p'
> refclock_oncore.c:204: warning: data definition has no type or storage 
> class
>     202         int     shmemfd;        /* Status shm descriptor */
>     203         pps_handle_t pps_h;
>     204         pps_params_t pps_p;
>
> I am wondering if this is the code I should be using?
>
> Any ideas?
>
Hmmm.  That Oncore driver seems to be sailing under a curse!!!

You'll need to file a bug report.  On the ntp web site, look for links 
to "Bugzilla".  Don't know what else to tell you; sorry about  that!