[LinuxPPS] ntp: add hardpps implementation

[Alexander Gordeev]

В Sun, 10 Oct 2010 17:29:30 +0200
Udo van den Heuvel <udovdh at xs4all.nl> пишет:

> On 2010-10-10 17:19, Remco dB wrote:
> > OK, 
> > 
> > I applied the latest Alex-patch on kernel-2.6.34 and got 'it' working 
> > according to Udo's howto on the PPS wiki. 
> > 
> > Interestingly, when I perform a ntpdc -c kern <LinuxPPS machine with hardpps> 
> > the PLL offset remains rock solid 0 sec. 
> > Status now is 0x2107 instead of 0x2007 or 0x2001.
> > 
> > However, when I do a ntpq -p the offset remains .. rock solid 127 us. 
> > When I do not use hardpps the system behaves more 'normal' and displays
> > with a ntpq -p query to an offset of +/- 0.001 on the same system.
> 
> Hmmm:
> [root at epia tor]# ntptime
> ntp_gettime() returns code 0 (OK)
>   time d05c5924.15be7f24  Sun, Oct 10 2010 17:24:20.084, (.084938024),
>   maximum error 7239 us, estimated error 4 us, TAI offset 0
> ntp_adjtime() returns code 0 (OK)
>   modes 0x0 (),
>   offset 0.000 us, frequency 3.669 ppm, interval 256 s,
>   maximum error 7239 us, estimated error 4 us,
>   status 0x2107 (PLL,PPSFREQ,PPSTIME,PPSSIGNAL,NANO),
>   time constant 4, precision 0.001 us, tolerance 500 ppm,
>   pps frequency 3.662 ppm, stability 0.013 ppm, jitter 4.543 us,
>   intervals 695, jitter exceeded 8, stability exceeded 0, errors 0.
> [root at epia tor]# ntpq -pn
>      remote           refid      st t when poll reach   delay   offset
> jitter
> ==============================================================================
> o127.127.20.0    .GPS.            0 l    6   16  377    0.000    0.006
>  0.002
> +194.109.22.18   193.79.237.14    2 u   15   64  377   21.115   -3.260
> 23.025
> +194.109.20.18   193.79.237.14    2 u   55   64  377   21.114   -3.681
> 34.720
> +193.67.79.202   .PPS.            1 u   24   64  377   21.120   -2.905
> 31.162
> +193.79.237.14   .PPS.            1 u   66   64  377   21.108   -3.685
> 31.010
> *83.163.219.98   .DCFa.           1 u   63   64  377   41.072   -9.059
>  6.893
>  134.221.205.12  .FLY.            1 u  169 1024    0    0.000    0.000
>  0.000
> 
> Any conclusions yet? (VIA Epia LT10000)
> 
> > PLL freq without hardpps is 157 ppm and with hardpps 30 ppm.
> 
> http://pindarots.xs4all.nl/mrtg/ntppll.html shows a lower pll number.
> 
> > I remember that there was an issue with CONFIG_NO_HZ. 
> 
> I have NO_HZ
> 
> > In my current hardpps kernel CONFIG_NO_HZ is set (y). 
> > Considering the 'rock solid' offset of 127 us, should I change this to 'n' ?
> 
> Any guru's?

To make hardpps() stable you have to disable CONFIG_NO_HZ if your
kernel version is greater or equal then 2.6.33. You should see it going
wild using ppstest or ppswatch from pps-tools. Can you please test it
and add this info to the wiki?

-- 
  Alexander
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: not available
Url : http://ml.enneenne.com/pipermail/linuxpps/attachments/20101010/e2ee3cd5/attachment.pgp