linux-decnet-user — Discussion for users of DECnet on Linux

Re: [Linux-decnet-user] LAT for linux using DECserver 90L+ and VT420

[John F. <jo...@fo...>]

On Tue, 30 Sep 2008, John Forkosh wrote:
> On Tue, 30 Sep 2008, Christine Caulfield wrote:
> [...]
>> The next thing to do would be to strace the latd daemon
>> (Warning, this generates a lot of output!) with timestamps
>> (strace -tt -p <pid>) to see just at which point the
>> keystroke is being delayed.
>> 
> Okay, I'll do that while entering keystrokes on the VT420,
> and then try to grep -v the  strace 2>output  to remove
> any unnecessary heartbeat-type stuff. If the strace output
> seems overly long, I'll put it on my website and post
> a link to the mailing list.

Full output is at http://www.forkosh.com/strace.txt

Here's what it represents:
   I typed ........ on the VT until a dot didn't echo.
Then I typed a / (after which ./ immediately echoed),
followed by more ...'s until the next dot didn't echo.
So you'll see .........../............/.....etc.
Each / means the dot before it didn't echo immediately.
And there are lots of timestamps in between.
I've reproduced a few snippets below, but don't know
if they're useful ones.

As mentioned previously, please enjoy your vacation,
and get around to this whenever you feel like it.
I'm in no rush and under no pressure whatsoever.
Thanks again,
John

A typical "dot snippet" is
----------------------------------------------------------------
19:24:58.846386 select(1024, [4 5 6], NULL, NULL, {0, 80000}) = 1
    (in [4], left{0, 80000})
19:24:58.846915 recvmsg(4, {msg_name(18)={sa_family=AF_PACKET,
    proto=0x6004, if2, pkttype=PACKET_HOST, addr(6)={1, 08002b337f19},
    msg_iov(1)=[{"\2\1\1\0\1\366n~
    \1\6\1\0.\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 1600}],
    msg_controllen=0, msg_flags=0}, 0) = 46
19:24:58.847324 write(7, ".", 1)        = 1
19:24:58.847656 ioctl(7, FIONREAD, [1]) = 0
19:24:58.847902 select(1024, [4 5 6], NULL, NULL, {0, 80000}) = 0
    (Timeout)
etc.

And here's a "dot snippet" that didn't echo (because the next
char I typed is a /)
----------------------------------------------------------------
19:24:58.846386 select(1024, [4 5 6], NULL, NULL, {0, 80000}) = 1
    (in [4], left{0, 80000})
19:24:58.846915 recvmsg(4, {msg_name(18)={sa_family=AF_PACKET,
    proto=0x6004, if2, pkttype=PACKET_HOST, addr(6)={1, 08002b337f19},
    msg_iov(1)=[{"\2\1\1\0\1\366n~
    \1\6\1\0.\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 1600}],
    msg_controllen=0, msg_flags=0}, 0) = 46
19:24:58.847324 write(7, ".", 1)        = 1
19:24:58.847656 ioctl(7, FIONREAD, [1]) = 0
19:24:58.847902 select(1024, [4 5 6], NULL, NULL, {0, 80000}) = 0
    (Timeout)
19:24:58.925342 read(7, ".", 126)       = 1
19:24:58.925610 sendto(4,
    "\0\1\1\366\1\0\177n\6\1\0\17\240\10\7\10osh@psi2star:
    ~"..., 46, 0, {sa_family=AF_PACKET, proto=0x6004, if2,
    pkttype=PACKET_MULTICAST,
    addr(6)={0, 08002b337f19}, 20) = 46
19:24:58.925975 gettimeofday({1222817098, 926074}, NULL) = 0
19:24:58.926194 time(NULL)              = 1222817098
19:24:58.926384 time(NULL)              = 1222817098
19:24:58.926588 select(1024, [4 5 6], NULL, NULL, {0, 80000}) = 0
    (Timeout)

Re: [Linux-decnet-user] LAT for linux using DECserver 90L+ and VT420

[John F. <jo...@fo...>]

On Tue, 30 Sep 2008, Christine Caulfield wrote:
>
> John Forkosh wrote:
>> I have a small thinwire soho lan with some VAXstations,
>> some PC's running (slackware 12.1) linux with DECnet and LAT
>> for linux, and a DECserver 90L+ with some VT420's.
>> The VT's connect to the VAXstations fine and dandy.
>> They also connect okay to the PC's, but with one small glitch.
>>       Every 11th or 12th or 13th or 14th (seems to vary
>> randomly in that range) keystroke isn't executed or echoed.
>> It seems to get buffered somewhere, though, because as soon
>> as I press another key, then both keystrokes are executed
>> and echoed immediately.  What's this behavior called (so that
>> I can google it which I've failed to do so far), and how
>> can I fix it (if it's fixable)?
>
> From the description it sounds like the keystroke echo is being buffered
> in the circuit timer buffer inside latd. Though that's odd because it's
> not specific to the 90L server. I do have a couple of 90L (not +)
> servers but I no longer have a power supply to drive them Then again,
> I'm also away from home at the moment so I can't do much off-hand anyway
> I'm afraid.
>
Thanks so much for the quick reply, Chrissie.  Hope you're
"away from home" having fun on vacation, in which case please
continue enjoying yourself until back at work.  This little
problem is only a minor annoyance to me -- definitely not worth
annoying anyone else about.

About the problem being "odd", I originally posted it to
comp.sys.dec and one of the replies, from Johnny Billquist,
says,
    "I've seen that too. I'm almost certain it is
     some kind of bug in the LAT implementation for Linux,
     but good luck in figuring that one out..."
So it may be odd, but apparently not unique.  Unfortunately,
he doesn't go into any detail about his configuration.

Another reply, from Bob Koehler, says,
    "We had similar problems with an early release of DECnet
     for Linux where the DECnet remote terminal protocol
     (not LAT) would drop characters.  Fixed it by updating
     to the latest patches to DECnet for Linux."
But this reply may be irrelevant to the problem at hand
(you'd know better than I would).
     By the way, I'm running dnprogs-2.43.tar.gz and
latd-1.25.tar.gz, though I see you've updated to
dnprogs-2.47.tar.gz and latd-1.27.tar.gz .  Your Release
Notes didn't seem (to me) to suggest anything related
to this behavior, but I'll be happy to upgrade before
running strace if you recommend it.

> The next thing to do would be to strace the latd daemon (Warning, this
> generates a lot of output!) with timestamps (strace -tt -p <pid>) to see
> just at which point the keystroke is being delayed.
>
Okay, I'll do that while entering keystrokes on the VT420,
and then try to grep -v the  strace 2>output  to remove
any unnecessary heartbeat-type stuff.

>> Thanks a lot,
>> John <jo...@fo...>
>> P.S.  Just subscribed to this list.  But can't seem to
>> find an archive for it.  Where's that at?
>
> The list archive URL is on the project home page, but it had a spurious
> full-stop in it, so it didn't work! I've fixed that, and the right place
> is here:
>
> http://sourceforge.net/mailarchive/forum.php?forum_name=linux-decnet-user
>
Got it.  Thanks a lot.

> Feel free to send the strace output to the list, but don't expect a fast
> response - they are hard work to search through, I'm frightfully busy
> (even when not away from home), and I haven't looked much at the latd
> code in a couple of years now ... sorry.
>
No problem.  And no hurry for me, anyway.  As above, please
enjoy yourself on vacation, and get around to this whenever
the spirit moves you.  If the strace output seems overly
long, I'll put it on my website and post a link to the
mailing list.

> LAT is pretty much a one-woman
> project unless you fancy digging into the code yourself!
>
I'm more of an application-level programmer, but will
be happy to take a peek if you can suggest where to begin,
i.e., hopefully somewhere deep down the structure chart,
above which there's probably no problem.
Thanks again,
John

Re: [Linux-decnet-user] LAT for linux using DECserver 90L+ and VT420

[Christine C. <Chr...@go...>]

John Forkosh wrote:
> I have a small thinwire soho lan with some VAXstations,
> some PC's running (slackware 12.1) linux with DECnet and LAT
> for linux, and a DECserver 90L+ with some VT420's.
> The VT's connect to the VAXstations fine and dandy.
> They also connect okay to the PC's, but with one small glitch.
>       Every 11th or 12th or 13th or 14th (seems to vary
> randomly in that range) keystroke isn't executed or echoed.
> It seems to get buffered somewhere, though, because as soon
> as I press another key, then both keystrokes are executed
> and echoed immediately.  What's this behavior called (so that
> I can google it which I've failed to do so far), and how
> can I fix it (if it's fixable)?

>From the description it sounds like the keystroke echo is being buffered
in the circuit timer buffer inside latd. Though that's odd because it's
not specific to the 90L server. I do have a couple of 90L (not +)
servers but I no longer have a power supply to drive them Then again,
I'm also away from home at the moment so I can't do much off-hand anyway
I'm afraid.

The next thing to do would be to strace the latd daemon (Warning, this
generates a lot of output!) with timestamps (strace -tt -p <pid>) to see
just at which point the keystroke is being delayed.


> Thanks a lot,
> John <jo...@fo...>
> P.S.  Just subscribed to this list.  But can't seem to
> find an archive for it.  Where's that at?
> 

The list archive URL is on the project home page, but it had a spurious
full-stop in it, so it didn't work! I've fixed that, and the right place
is here:

http://sourceforge.net/mailarchive/forum.php?forum_name=linux-decnet-user

Feel free to send the strace output to the list, but don't expect a fast
response - they are hard work to search through, I'm frightfully busy
(even when not away from home), and I haven't looked much at the latd
code in a couple of years now ... sorry. LAT is pretty much a one-woman
project unless you fancy digging into the code yourself!

Chrissie

[Linux-decnet-user] LAT for linux using DECserver 90L+ and VT420

[John F. <jo...@fo...>]

I have a small thinwire soho lan with some VAXstations,
some PC's running (slackware 12.1) linux with DECnet and LAT
for linux, and a DECserver 90L+ with some VT420's.
The VT's connect to the VAXstations fine and dandy.
They also connect okay to the PC's, but with one small glitch.
      Every 11th or 12th or 13th or 14th (seems to vary
randomly in that range) keystroke isn't executed or echoed.
It seems to get buffered somewhere, though, because as soon
as I press another key, then both keystrokes are executed
and echoed immediately.  What's this behavior called (so that
I can google it which I've failed to do so far), and how
can I fix it (if it's fixable)?
Thanks a lot,
John <jo...@fo...>
P.S.  Just subscribed to this list.  But can't seem to
find an archive for it.  Where's that at?

[Linux-decnet-user] dnprogs 2.47 released

[Christine C. <Chr...@go...>]

dnprogs 2.47 is a small update to 2.46

* Fix libdnet_daemon so that dnet_accept & dnet_reject work on
    big-endian systems. This fixes a lot of rejected connections from RSX!
* nml now shows number of active links with nodes list
* nml now supports "show known links" & "show known objects"
* dnetstat shows object names instead of numbers, unless you specify -n

Download it from
http://sourceforge.net/project/showfiles.php?group_id=4993&package_id=5032&release_id=627898

-- 
Chrissie

[Linux-decnet-user] dnprogs 2.46 for Linux released (NEW FEATURES!)

[Christine C. <chr...@go...>]

I'm very pleased to announce the new dnprogs release. With not one, not
two, but FOUR new programs - you lucky people :-)

The first two were contributed by Philipp 'ph3-der-loewe' Schafft and
consist of two command-line utilities:

dnetstat is a version of netstat for DECnet, it reformats the /proc file
output in a more user-friendly manner.
dnetcat is a DECnet version of netcat, an invaluable utility for all
network nerds. Philipp has offered to update the wiki to show how to use
this utility to run SSH over DECnet!

And finally, I have also been busy. I have done something I said I'd
write years ago and that's an Network Management Listener program. Yes,
finally you can go into NCP on a VMS or RSX system (yes, I tested it on
RSX, thanks Johnny) and type in "TELL LINUX SHOW ADJACENT NODES". It
only supports a subset of commands at the moment, but I will probably
add more. It currently will do:

SHOW/LIST KNOWN NODES
SHOW ADJACENT NODES
SHOW EXEC CHAR

To make this work you'll need to add a new line to dnetd.conf - see the
sample file shipped with the package for the details.

There's also a client side version of that, well partly. dncopynodes
allows a Linux user to pull a node list off a remote VMS or RSX system
and build up a full decnet.conf file. It will fill in the current exec
details too so that the file it generates can simply be copied to /etc.
WARNING ... DO NOT do 'dncopynodes > /etc/decnet.conf' as it will
overwrite the file before it gets a chance to use it to make the initial
connection!


http://sourceforge.net/project/showfiles.php?group_id=4993&package_id=5032&release_id=624716

-- 
Chrissie