160

Unix/Linux 网络时间协议版本3 Network Time Protocol Version 3 (NTP) distribution for Unix systems

Replied: Mon, 07 Jul 1997 20:24:24 -0400
Replied: "Damon at work <[e-mail]> "
Replied: Mon, 07 Jul 1997 18:32:41 -0400
Replied: "Damon at work <[e-mail]> [e-mail]"
Return-Path: [e-mail] 
Return-Path: <[e-mail]>
Received: from mailgate.exnet.com ([204.137.193.226])
	by whimsy.udel.edu (8.8.5/8.8.5) with SMTP id JAA08336
	for <[e-mail]>; Mon, 7 Jul 1997 09:44:50 GMT
Received: from exnet.com (assam.exnet.com) by mailgate.exnet.com with SMTP id AA03149
	(5.67a/IDA-1.4.4 for [e-mail]); Mon, 7 Jul 1997 10:44:12 +0100
Received: from maildrop.exnet.com (oolong.exnet.com) by exnet.com with SMTP id AA10527
	(5.67a/IDA-1.4.4); Mon, 7 Jul 1997 10:43:46 +0100
From: Damon at work <[e-mail]>
Received: by maildrop.exnet.com (SMI-8.6/client-1.4DHD)
	id KAA16049; Mon, 7 Jul 1997 10:43:43 +0100
Date: Mon, 7 Jul 1997 10:43:43 +0100
Message-Id: <[e-mail]>
To: [e-mail], [e-mail]
Subject: Updates to ARCRON clock driver.
Cc: [e-mail]
Content-Type: X-sun-attachment

----------
X-Sun-Data-Type: text
X-Sun-Data-Description: text
X-Sun-Data-Name: text
X-Sun-Charset: us-ascii
X-Sun-Content-Lines: 57

Hi folks,

Some updates to the driver:

 1) About half the memory footprint (most of my debugging code
    disabled by default).

 2) Some general tweaks and improvements, including an attempt
    to increase sampling frequency at the point of a potential
    leap-second action (though this did not seem to work on 30th June,
    so it is disabled for now).

 3) The code avoids the clock tty becoming the controlling tty.

Diffs from the 5.90.2 code are supplied.

Harlan, I believe I have preserved your changes for older compilers.

BTW, we don't seemed to have turned up in the copyright document yet.
And I was staking my reputation on it...

Regards,

Damon

BTW, are you aware that (at least under 3.50[.0]):

 a) The current PARSE install/build stuff is way out of sync with
    your new build structure (it still talks about the 3.3/4.3 style
    build flags).

 b) Same with the kernel/ stuff.

 c) They ask you to build stuff into the kernel, and won't build the
    PARSE stuff unless you have copied stuff into the sys area, but
    that is unnecessary (and therefore a BadThing(TM) IMHO) under
    Solaris-2.

 d) The PARSE module for Solaris-2 seems to blow the kernel (2.5/2.5.1)
    up immediately (and at least one other poster on c.p.t.n had the
    same experience).

 e) I think ppsclock paniced my kernel too, but I was getting confused
    at that stage and gave up, since I did not want to modify my sys
    area unnecessarily, and got fed up with hacking the config stuff.

I only report some of these to you rather than Frank, since some of the
pertinent docs have Dave's name on rather than Frank's.

I have had much more success in the past with PARSE and indeed have
written STREAMS modules under SunOS, so I'm sure I can fix all this
when I get a spare month...  B^>

Would you like me to mail Frank directly with this, or would one of you
folks like to pass this on to him?

Damon
----------
X-Sun-Data-Type: default
X-Sun-Data-Description: default
X-Sun-Data-Name: driver27_DIFF
X-Sun-Charset: us-ascii
X-Sun-Content-Lines: 21

24,25c24,25
< <p>This documentation describes version V1.1 (1997/06/23) of the source
< and has been tested (amongst others) against xntpd3-5.90 on Solaris-1 (SunOS 4.1.3_U1 on
---
> <p>This documentation describes version V1.0 (1997/05/21) of the source
> and has been tested against xntpd3-5.90 on Solaris-1 (SunOS 4.1.3_U1 on
33,35d32
< <p>This code has been significantly slimmed down since the V1.0 version,
< roughly halving the memory footprint of its code and data.
< 
323,332d319
< <dt><code>ARCRON_LEAPSECOND_KEEN</code></dt>
< <dd>When is defined, we try to do a resync to MSF as soon as possible
< in the first hour of the morning of the first day of the first and
< seventh months, ie just after a leap-second insertion or deletion would
< happen if it is going to.  This should help compensate for the fact
< that this clock does not continuously sample MSF, which compounds the
< fact that MSF itself gives no warning of an impending leap-second
< event.  This code did not seem functional at the leap-second insertion
< of 30th June 1997 so is by default disabled.
< 
----------
X-Sun-Data-Type: default
X-Sun-Data-Description: default
X-Sun-Data-Name: refclock_arc_DIFF
X-Sun-Charset: us-ascii
X-Sun-Content-Lines: 358

10d9
<     static const char arc_version[] = { "V1.1 1997/06/23" };
12c11
< #undef ARCRON_DEBUG /* Define only while in development... */
---
>     static const char arc_version[] = { "V1.0 1997/05/21" };
19,20c18
< #ifndef ARCRON_OWN_FILTER
< #undef ARCRON_OWN_FILTER /* Use own median filter only for versions before 3-5.90.1. */
---
> #define ARCRON_OWN_FILTER 1 /* Use own median filter to get round 3-5.90 bug. */
22d19
< #endif
28,33d24
< #ifndef ARCRON_NOT_LEAPSECOND_KEEN
< #ifndef ARCRON_LEAPSECOND_KEEN
< #undef ARCRON_LEAPSECOND_KEEN /* Respond quickly to leap seconds: doesn't work yet. */
< #endif
< #endif
< 
78,81d68
< MAJOR CHANGES SINCE V1.0
< ========================
<  1) Removal of pollcnt variable that made the clock go permanently
<     off-line once two time polls failed to gain responses.
83,92d69
<  2) Avoiding (at least on Solaris-2) terminal becoming the controlling
<     terminal of the process when we do a low-level open().
< 
<  3) Additional logic (conditional on ARCRON_LEAPSECOND_KEEN being
<     defined) to try to resync quickly after a potential leap-second
<     insertion or deletion.
< 
<  4) Code significantly slimmer at run-time than V1.0.
< 
< 
217a195,197
>   * Avoid /dev/arcX becoming our controlling terminal.  Do we really
>     want a ^C or somesuch on the tty line aborting xntpd (would it)?
> 
227c207
<     correction would be disabled by default).
---
>     correction would be disabled by defaulr).
442,447d421
< /* Flags for a raw open() of the clock serial device. */
< #ifdef O_NOCTTY /* Good, we can avoid tty becoming controlling tty. */
< #define OPEN_FLAGS (O_RDWR | O_NOCTTY)
< #else           /* Oh well, it may not matter... */
< #define OPEN_FLAGS (O_RDWR)
< #endif
449d422
< 
470a444
>     int pollcnt;        /* Poll message counter. */
487,489d460
< #ifdef ARCRON_LEAPSECOND_KEEN
<     /* The flag `possible_leap' is set non-zero when any MSF unit
<        thinks a leap-second may have happened.
491,511d461
<        Set whenever we receive a valid time sample in the first hour of
<        the first day of the first/seventh months.
< 
<        Outside the special hour this value is unconditionally set
<        to zero by the receive routine.
< 
<        On finding itself in this timeslot, as long as the value is
<        non-negative, the receive routine sets it to a positive value to
<        indicate a resync to MSF should be performed.
< 
<        In the poll routine, if this value is positive and we are not
<        already resyncing (eg from a sync that started just before
<        midnight), start resyncing and set this value negative to
<        indicate that a leap-triggered resync has been started.  Having
<        set this negative prevents the receive routine setting it
<        positive and thus prevents multiple resyncs during the witching
<        hour.
<      */
<     static int possible_leap = 0;       /* No resync required by default. */
< #endif
< 
553d502
< #ifdef ARCRON_DEBUG
555d503
< #endif
576c524
< #ifdef ARCRON_DEBUG
---
> 
578d525
< #endif
586,592c533,534
<         if(c != '\0') {
<             if(write(pp->io.fd, &c, 1) != 1) {
<                 syslog(LOG_NOTICE, "ARCRON: write to fd %d failed", pp->io.fd);
<             }
< #ifdef ARCRON_DEBUG
<         else if(debug) { printf("arc: sent `%2.2x', fd %d.\n", c, pp->io.fd); }
< #endif
---
>         if((c != '\0') && (write(pp->io.fd, &c, 1) != 1)) {
>             syslog(LOG_NOTICE, "ARCRON: could not write to fd %d", pp->io.fd);
593a536,539
> 
>         if(debug && (c != '\0'))
>             { printf("arc: sending `%2.2x' on fd %d.\n", c, pp->io.fd); }
> 
608a555
>         int i;
614d560
< #ifdef ARCRON_DEBUG
618d563
< #endif
628d572
< #ifdef ARCRON_DEBUG
630d573
< #endif
632,634c575
< #ifdef DEBUG
<             if(debug) { printf("arc: failed [TTYCLK] to open %s.\n", device); }
< #endif
---
>             if(debug) { printf("Failed [TTYCLK] to open %s.\n", device); }
638d578
< #ifdef ARCRON_DEBUG
640,641c580
< #endif
<         fd = open(device, OPEN_FLAGS);
---
>         fd = open(device, O_RDWR);
643,645c582
< #ifdef DEBUG
<             if(debug) { printf("arc: failed [open()] to open %s.\n", device); }
< #endif
---
>             if(debug) { printf("Failed [open()] to open %s.\n", device); }
650c587
< #ifdef ARCRON_DEBUG
---
> 
652,653c589
<             { printf("Opened RS232 port with file descriptor %d.\n", fd); }
< #endif
---
>             { printf("Opening RS232 port with file descriptor %d.\n", fd); }
678d613
<         /* Set structure to all zeros... */
695,696d629
<         /* Spread out resyncs so that they should remain separated. */
<         up->next_resync = current_time + INITIAL_RESYNC_DELAY + (67*unit)%1009;
698,700c631
< #if 0 /* Not needed because of zeroing of arcunit structure... */
<         up->resyncing = 0;              /* Not resyncing yet. */
<         up->saved_flags = 0;            /* Default is all flags off. */
---
>         up->pollcnt = 2;
702,703d632
<         {
<         int i;
705,707d633
<         }
< #endif
< 
715a642
>         up->ev.next = up->ev.prev = 0;  /* Be ultra-cautious... */
716a644,648
> 
>         /* Spread out resyncs so that they should remain separated. */
>         up->next_resync = current_time + INITIAL_RESYNC_DELAY + (67*unit)%1009;
>         up->resyncing = 0;              /* Not resyncing yet. */
>         up->saved_flags = 0;            /* Default is all flags off. */
767c699
< #ifdef ARCRON_DEBUG
---
> 
769,771c701,702
< #endif
<     if(spaceleft < sl) { /* Should not normally happen... */
< #ifdef ARCRON_DEBUG
---
> 
>     if(spaceleft < sl) {
774d704
< #endif
825d754
< #ifdef DEBUG
827d755
< #endif
870a799
> 
872c801
< #ifdef DEBUG
---
> 
877d805
< #endif
890d817
< #ifdef ARCRON_DEBUG
895d821
< #endif
902c828
< #ifdef ARCRON_DEBUG
---
> 
913d838
< #endif
931d855
< #ifdef ARCRON_DEBUG
936a861
> 
944d868
< #endif
955d878
< #ifdef ARCRON_DEBUG
958d880
< #endif
980d901
< #ifdef ARCRON_DEBUG
982d902
< #endif
995d914
< #ifdef ARCRON_DEBUG
997d915
< #endif
1002d919
< #ifdef DEBUG
1004d920
< #endif
1006d921
< #ifdef DEBUG
1013d927
< #endif
1043d956
< #ifdef ARCRON_DEBUG
1045d957
< #endif
1049d960
< #ifdef ARCRON_DEBUG
1051d961
< #endif
1055a966,969
> 
>         /* Give us some leeway before deciding that the clock is ill... */
>         up->pollcnt = 2;
> 
1077,1078d990
< #ifdef ARCRON_DEBUG
<             /* Would expect to have caught major problems already... */
1080d991
< #endif
1121c1032
< #ifdef DEBUG
---
> 
1128d1038
< #endif
1146d1055
<                 up->status = status;
1182d1090
< #ifdef ARCRON_DEBUG