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

RE: [Linux-decnet-user] protocols

[Colin B. <col...@xd...>]

There's a lot of Phase IV and Phase V information in the VMS manuals, FAQ
and Ask the Wizard sections at www.openvms.compaq.com

There's some DECnet Phase IV to V migration and implementation information
at www.xdelta.co.uk - I wrote this a while back as part of a DECUS (Digital
Equipment Computer Users Society) session.

There's lots of good stuff at
http://www.research.compaq.com/wrl/DECarchives/DTJ/ - an archive of (some
of) the old Digital Technical Journals.

There's also a VMS webring which can bring up some good stuff too -
http://www.webring.org/cgi-bin/webring?ring=openvms&id=44&list

You have the pointer to the DECnet Phase IV specifications. The Phase V
specifications aren't (to my knowledge) yet published on the web, nor are
the LAT specifications (LAT is a layer 2 protocol used for serial traffic
between terminal servers and host systems - it's a completely separate
protocol to DECnet).

Hope this helps,
Colin.
  -----Original Message-----
  From: lin...@li...
[mailto:lin...@li...]On Behalf Of Brendan
Minchin
  Sent: Sunday, April 28, 2002 3:39 AM
  To: lin...@li...
  Subject: [Linux-decnet-user] protocols


  Can you please tell me the layers of the DECnet protocol and compare them
to the OSI seven layer model? Also can you tell me where to get info about
each layer in the DECnet?
  Cheers
  Brendan

Re: [Linux-decnet-user] protocols

[Paul K. <pk...@eq...>]

Excerpt of message (sent 28 April 2002) by Brendan Minchin:

> Can you please tell me the layers of the DECnet protocol and compare
> them to the OSI seven layer model? Also can you tell me where to get
> info about each layer in the DECnet?

I meant to answer that first question too...

Decnet phase 4 predates the OSI model.  Phase 5 had a number of its
layer names changed so they wouldn't be as confusing to someone who
reads the OSI model.  Apart from that, the two have basically the same
layering:

Osi	Phase4		Phase5
7	apps		apps
6	-		-
5	(NSP)		Session Ctl
4	NSP		Transport
3	Transport	Routing
2	Data Link	Data Link
1	(data link)	Physical

So there's a pretty close match with two or three exceptions.  Phase 4
does not have a separate physical layer nor a separate transport
layer; you can see the functions of these two as part of the data link
and transport layers, respectively.  There is no layer 6 -- but then
again, I don't think there has ever been a layer 6 in recorded
history.  Finally, the names are different, which isn't all that bad
except for layer 3, which confusingly is called "transport" in Phase
4.  (Layer 4 is "NSP" which if you expand it also becomes confusing:
"Network services protocol".  So just keep it as "NSP"!)

	 paul

Re: [Linux-decnet-user] protocols

[Paul K. <pk...@eq...>]

Excerpt of message (sent 28 April 2002) by Brendan Minchin:
> Can you please tell me the layers of the DECnet protocol and compare
> them to the OSI seven layer model? Also can you tell me where to get
> info about each layer in the DECnet?

You can find the (phase IV) specs at
  http://gatekeeper.dec.com/pub/DEC/DECnet/PhaseIV/

      paul

[Linux-decnet-user] protocols

[Brendan M. <cre...@wn...>]

Can you please tell me the layers of the DECnet protocol and compare =
them to the OSI seven layer model? Also can you tell me where to get =
info about each layer in the DECnet?
Cheers
Brendan

RE: [Linux-decnet-user] Closing DECNET connections

[Lhota R. <rl...@co...>]

Hi,

thank you for the patch. I appied it, recompilled the kernel, but
unfortunatelly, it doesn't work for me. During dnping execution, the
program tries to connect to the destination vax machine and sometimes
this connect takes very long time or freezes at all and then dnping ends
with timeout and the RUN IMMED connection still remains forever.

Robert

> -----Original Message-----
> From: Patrick Caulfield [mailto:pa...@ty...]
> Sent: Thursday, April 18, 2002 4:59 PM
> To: Lin...@li...
> Subject: Re: [Linux-decnet-user] Closing DECNET connections
>=20
>=20
> On Thu, Apr 18, 2002 at 02:51:52PM +0100, Steven Whitehouse wrote:
> > Hi,
> >=20
> > >=20
> > Patrick says that it happens in 2.4.0, so you could try=20
> something older than
> > that, but I certainly wouldn't recommend using it for=20
> production at all. It
> > seems that there is a real bug here, but I'm unfortunately=20
> rather short of
> > time to investigate it right now.
> >=20
> > I wonder if its actually a DECnet system call that gets=20
> interrupted in
> > such a way as to get a reference count wrong during the=20
> exit code path
> > and thus the socket never gets destroyed.
> >=20
> > Its odd because when the process exits, the release method=20
> should be called
> > for the socket and that has all the socket shut down code=20
> in it so whatever
> > happens it should work, though I know it doesn't,
> >=20
>=20
> Here's a patch which works for me:
>=20
> $ diff -u net/decnet/af_decnet.c.orig net/decnet/af_decnet.c
> --- net/decnet/af_decnet.c.orig Thu Apr 18 15:46:11 2002
> +++ net/decnet/af_decnet.c      Thu Apr 18 15:46:34 2002
> @@ -747,10 +747,10 @@
>         struct sock *sk =3D sock->sk;
>=20
>         if (sk) {
> -               sock_orphan(sk);
>                 sock_hold(sk);
>                 lock_sock(sk);
>                 dn_destroy_sock(sk);
> +               sock_orphan(sk);
>                 release_sock(sk);
>                 sock_put(sk);
>         }
>=20
> I'll forward it to Dave Miller for inclusion in the stock kernel.
>=20
> patrick
>=20
>=20
> _______________________________________________
> Linux-decnet-user mailing list
> Lin...@li...
> https://lists.sourceforge.net/lists/listinfo/linux-decnet-user
>=20

Re: [Linux-decnet-user] Closing DECNET connections

[Patrick C. <pa...@ty...>]

On Thu, Apr 18, 2002 at 02:51:52PM +0100, Steven Whitehouse wrote:
> Hi,
> 
> > 
> Patrick says that it happens in 2.4.0, so you could try something older than
> that, but I certainly wouldn't recommend using it for production at all. It
> seems that there is a real bug here, but I'm unfortunately rather short of
> time to investigate it right now.
> 
> I wonder if its actually a DECnet system call that gets interrupted in
> such a way as to get a reference count wrong during the exit code path
> and thus the socket never gets destroyed.
> 
> Its odd because when the process exits, the release method should be called
> for the socket and that has all the socket shut down code in it so whatever
> happens it should work, though I know it doesn't,
> 

Here's a patch which works for me:

$ diff -u net/decnet/af_decnet.c.orig net/decnet/af_decnet.c
--- net/decnet/af_decnet.c.orig Thu Apr 18 15:46:11 2002
+++ net/decnet/af_decnet.c      Thu Apr 18 15:46:34 2002
@@ -747,10 +747,10 @@
        struct sock *sk = sock->sk;

        if (sk) {
-               sock_orphan(sk);
                sock_hold(sk);
                lock_sock(sk);
                dn_destroy_sock(sk);
+               sock_orphan(sk);
                release_sock(sk);
                sock_put(sk);
        }

I'll forward it to Dave Miller for inclusion in the stock kernel.

patrick

Re: [Linux-decnet-user] Closing DECNET connections

[Steven W. <st...@gw...>]

Hi,

> 
> Hi,
> 
> >It looks from your description as if that is very much the case. I'm
> sure
> >that this didn't use to happen as surely someone (me even!) would have
> >noticed it before. I've been rather busy lately and not had the time to
> >check this bug out properly. At least I can't see why its happening
> from
> >the code.
> 
> I would like to try some older version where the problem doesn't occur.
> Can you
> tell me which version should I try?
> 
> Thanks,
> Robert
> 
Patrick says that it happens in 2.4.0, so you could try something older than
that, but I certainly wouldn't recommend using it for production at all. It
seems that there is a real bug here, but I'm unfortunately rather short of
time to investigate it right now.

I wonder if its actually a DECnet system call that gets interrupted in
such a way as to get a reference count wrong during the exit code path
and thus the socket never gets destroyed.

Its odd because when the process exits, the release method should be called
for the socket and that has all the socket shut down code in it so whatever
happens it should work, though I know it doesn't,

Steve.

RE: [Linux-decnet-user] Closing DECNET connections

[Lhota R. <rl...@co...>]

Hi,

>It looks from your description as if that is very much the case. I'm
sure
>that this didn't use to happen as surely someone (me even!) would have
>noticed it before. I've been rather busy lately and not had the time to
>check this bug out properly. At least I can't see why its happening
from
>the code.

I would like to try some older version where the problem doesn't occur.
Can you
tell me which version should I try?

Thanks,
Robert



-----Original Message-----
From: Steven Whitehouse [mailto:st...@gw...]
Sent: Thursday, April 18, 2002 10:41 AM
To: Lhota Robert
Cc: Lin...@li...
Subject: Re: [Linux-decnet-user] Closing DECNET connections


Hi,

>=20
>=20
> > If you do that command on VMS then the Linux connections will
timeout.
> >=20
> The problem is, that I unfortunatelly dont have access to VAX/VMS
> machines at all because they belong to our customer (I am implementing
> part of SLM and all I have are DECNET addresses of VAX machines). So
my
> only possibility is to clear connections from Linux, which is
impossible
> or only possible by rebooting the machine. Or am I wrong?
>=20
> Robert
>=20
>=20
It looks from your description as if that is very much the case. I'm
sure
that this didn't use to happen as surely someone (me even!) would have
noticed it before. I've been rather busy lately and not had the time to
check this bug out properly. At least I can't see why its happening from
the code.

If someone knows that it never happens on a kernel of a particular
version
please post to the list (or just to me personally) as that will no doubt
give me a bit clue as to what the problem is.

The way that sockets shut down is completely independant to whether a
signal
has been received by the process or not. Thats the odd thing. I don't
see
why its a problem only when signals have been involved,

Steve.

Re: [Linux-decnet-user] Closing DECNET connections

[Patrick C. <pa...@ty...>]

On Thu, Apr 18, 2002 at 10:50:44AM +0200, Lhota Robert wrote:
> 
> > If you do that command on VMS then the Linux connections will timeout.
> > 
> The problem is, that I unfortunatelly dont have access to VAX/VMS
> machines at all because they belong to our customer (I am implementing
> part of SLM and all I have are DECNET addresses of VAX machines). So my
> only possibility is to clear connections from Linux, which is impossible
> or only possible by rebooting the machine. Or am I wrong?

You need to reboot the machine I'm afraid. the init.d script only restarts the
daemons. The kernel module is currently unloadable.

Once the machine has rebooted it will take a while for VMS to tidy out the dead
connections but they will go away eventually.


patrick

Re: [Linux-decnet-user] Closing DECNET connections

[Steven W. <st...@gw...>]

Hi,

> 
> 
> > If you do that command on VMS then the Linux connections will timeout.
> > 
> The problem is, that I unfortunatelly dont have access to VAX/VMS
> machines at all because they belong to our customer (I am implementing
> part of SLM and all I have are DECNET addresses of VAX machines). So my
> only possibility is to clear connections from Linux, which is impossible
> or only possible by rebooting the machine. Or am I wrong?
> 
> Robert
> 
> 
It looks from your description as if that is very much the case. I'm sure
that this didn't use to happen as surely someone (me even!) would have
noticed it before. I've been rather busy lately and not had the time to
check this bug out properly. At least I can't see why its happening from
the code.

If someone knows that it never happens on a kernel of a particular version
please post to the list (or just to me personally) as that will no doubt
give me a bit clue as to what the problem is.

The way that sockets shut down is completely independant to whether a signal
has been received by the process or not. Thats the odd thing. I don't see
why its a problem only when signals have been involved,

Steve.

RE: [Linux-decnet-user] Closing DECNET connections

[Lhota R. <rl...@co...>]

> If you do that command on VMS then the Linux connections will timeout.
>=20
The problem is, that I unfortunatelly dont have access to VAX/VMS
machines at all because they belong to our customer (I am implementing
part of SLM and all I have are DECNET addresses of VAX machines). So my
only possibility is to clear connections from Linux, which is impossible
or only possible by rebooting the machine. Or am I wrong?

Robert



> -----Original Message-----
> From: Patrick Caulfield [mailto:pa...@ty...]
> Sent: Wednesday, April 17, 2002 4:44 PM
> To: Lin...@li...
> Subject: Re: [Linux-decnet-user] Closing DECNET connections
>=20
>=20
> On Wed, Apr 17, 2002 at 04:27:49PM +0200, Lhota Robert wrote:
> > Thank you very much for the timeout option. I think it solves my
> > problem. I had to comment out the section:
> >=20
> > if (options & DNF_TIMEOUT)
> > 	{
> > 	    int status;
> > 	    fd_set in_fd;
> >=20
> > 	    FD_ZERO(&in_fd);=20
> > ...
> >=20
> > because it caused the program exit with "Timeout" every=20
> time when used
> > with options -t -w together (f.e. dnping -t -w 10 11.42). Instead of
> > that I extended the alarm to the all program by commenting=20
> out alarm(0)
> > function. I am not c-coder at all but this works for me now.
>=20
> Aha typo! here's the fix:
> diff -u -r1.4 dnping.c
>=20
> --- dnping.c    17 Apr 2002 07:42:31 -0000      1.4
> +++ dnping.c    17 Apr 2002 14:41:19 -0000
> @@ -423,7 +423,7 @@
>             timeout.tv_sec =3D timeout_sec;
>             timeout.tv_usec =3D 0;
> =20
> -           status =3D select(sockfd+1, &in_fd, NULL, NULL, &tv);
> +           status =3D select(sockfd+1, &in_fd, NULL, NULL, &timeout);
>             if (status < 0)
>             {
>                 perror("select");
>=20
>=20
> =20
> > Do I understand it right, that
> > mc ncp sho kno links
> > and
> > NCP DISCONECT LINK=20
> > are VAX machine commands and cannot be used on Linux=20
> implementation of
> > DECNET? Is there than any possibility how to clear the "zombie
> > connections" from Linux? I tried to restart decnet vie=20
> startup scripts
> > (/etc/init.d/decnet on RedHat) but it didn't clear the=20
> connections and
> > even than  I could not dnping the hosts that were in the=20
> connections any
> > more (after decnet restart).
>=20
> Yes, those are VMS commands, there is no Linux equivalent=20
> that I know of - and
> you need privileges on VMS to do it.
>=20
> If you do that command on VMS then the Linux connections will timeout.
>=20
> patrick
>=20
>=20
> _______________________________________________
> Linux-decnet-user mailing list
> Lin...@li...
> https://lists.sourceforge.net/lists/listinfo/linux-decnet-user
>=20

Re: [Linux-decnet-user] Closing DECNET connections

[Patrick C. <pa...@ty...>]

On Wed, Apr 17, 2002 at 04:27:49PM +0200, Lhota Robert wrote:
> Thank you very much for the timeout option. I think it solves my
> problem. I had to comment out the section:
> 
> if (options & DNF_TIMEOUT)
> 	{
> 	    int status;
> 	    fd_set in_fd;
> 
> 	    FD_ZERO(&in_fd); 
> ...
> 
> because it caused the program exit with "Timeout" every time when used
> with options -t -w together (f.e. dnping -t -w 10 11.42). Instead of
> that I extended the alarm to the all program by commenting out alarm(0)
> function. I am not c-coder at all but this works for me now.

Aha typo! here's the fix:
diff -u -r1.4 dnping.c

--- dnping.c    17 Apr 2002 07:42:31 -0000      1.4
+++ dnping.c    17 Apr 2002 14:41:19 -0000
@@ -423,7 +423,7 @@
            timeout.tv_sec = timeout_sec;
            timeout.tv_usec = 0;
 
-           status = select(sockfd+1, &in_fd, NULL, NULL, &tv);
+           status = select(sockfd+1, &in_fd, NULL, NULL, &timeout);
            if (status < 0)
            {
                perror("select");


 
> Do I understand it right, that
> mc ncp sho kno links
> and
> NCP DISCONECT LINK 
> are VAX machine commands and cannot be used on Linux implementation of
> DECNET? Is there than any possibility how to clear the "zombie
> connections" from Linux? I tried to restart decnet vie startup scripts
> (/etc/init.d/decnet on RedHat) but it didn't clear the connections and
> even than  I could not dnping the hosts that were in the connections any
> more (after decnet restart).

Yes, those are VMS commands, there is no Linux equivalent that I know of - and
you need privileges on VMS to do it.

If you do that command on VMS then the Linux connections will timeout.

patrick

RE: [Linux-decnet-user] Closing DECNET connections

[Lhota R. <rl...@co...>]

Thank you very much for the timeout option. I think it solves my
problem. I had to comment out the section:

if (options & DNF_TIMEOUT)
	{
	    int status;
	    fd_set in_fd;

	    FD_ZERO(&in_fd);=20
...

because it caused the program exit with "Timeout" every time when used
with options -t -w together (f.e. dnping -t -w 10 11.42). Instead of
that I extended the alarm to the all program by commenting out alarm(0)
function. I am not c-coder at all but this works for me now.

Do I understand it right, that
mc ncp sho kno links
and
NCP DISCONECT LINK=20
are VAX machine commands and cannot be used on Linux implementation of
DECNET? Is there than any possibility how to clear the "zombie
connections" from Linux? I tried to restart decnet vie startup scripts
(/etc/init.d/decnet on RedHat) but it didn't clear the connections and
even than  I could not dnping the hosts that were in the connections any
more (after decnet restart).

Robert Lhota


-----Original Message-----
From: Patrick Caulfield [mailto:pa...@ty...]
Sent: Wednesday, April 17, 2002 9:46 AM
To: Lin...@li...
Subject: Re: [Linux-decnet-user] Closing DECNET connections


I've added a timeout option to dnping which may help ease your problems.

Get it from CVS or=20

http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/linux-decnet/dnprogs/apps
/dnping.c?rev=3D1.4&content-type=3Dtext


patrick


_______________________________________________
Linux-decnet-user mailing list
Lin...@li...
https://lists.sourceforge.net/lists/listinfo/linux-decnet-user

Re: [Linux-decnet-user] Closing DECNET connections

[Patrick C. <pa...@ty...>]

I've added a timeout option to dnping which may help ease your problems.

Get it from CVS or 

http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/linux-decnet/dnprogs/apps/dnping.c?rev=1.4&content-type=text


patrick

Re: [Linux-decnet-user] Closing DECNET connections

[Patrick C. <pa...@ty...>]

On Tue, Apr 16, 2002 at 02:29:04PM +0100, Steven Whitehouse wrote:
> Hi,
> 
> > 
> Next thing is to strace. Perhaps escaping from one of the loops in
> sendmsg() or recvmsg() gets some reference counting wrong ? At least
> thats my best guess currently. I'll have a closer look later on.

Only strace exits if I hit ^C :(
 
> > 
> > P.S. Just noticed this too - I think It'd be nice if the link numbers in
> > /proc/net/decnet was in decimal rather than hex then it is possible to match
> > them up with those shown by VMS.
> > 
> Perhaps we could do that in 2.5. I'd prefer not to change things in the stable
> version unless they really cause people problems,

Fair comment.

patrick

Re: [Linux-decnet-user] Closing DECNET connections

[Steven W. <st...@gw...>]

Hi,

> 
> I've never seen it before but now it seems consistent, if I ^C a process with an
> open connection it stays in RUN. I'm running kernel 2.4.17 
> 
> Oddly adding the signal handler makes no difference - the sockets still stick in
> RUN (yes, I put a prinf in the signal handler so I knew it was being called!)
> 
Ok. Well that means I'm now more convinced than I was that its a kernel bug.

> But they go away if I let the program run to completion.
> 
> Wierd.
> 
> patrick
> 
Next thing is to strace. Perhaps escaping from one of the loops in
sendmsg() or recvmsg() gets some reference counting wrong ? At least
thats my best guess currently. I'll have a closer look later on.

> 
> P.S. Just noticed this too - I think It'd be nice if the link numbers in
> /proc/net/decnet was in decimal rather than hex then it is possible to match
> them up with those shown by VMS.
> 
Perhaps we could do that in 2.5. I'd prefer not to change things in the stable
version unless they really cause people problems,

Steve.

Re: [Linux-decnet-user] Closing DECNET connections

[Patrick C. <pa...@ty...>]

On Tue, Apr 16, 2002 at 02:06:00PM +0100, Steven Whitehouse wrote:
> Hi,
> 
> [snip]
> > 
> > 
> > Steve?
> > 
> > patrick
> > 
> > 
> Hmm. Not sure about that. When the socket gets closed (along with all the
> other fds of that process it should get shutdown. The release function has
> the code in it to force that. Are you sure that there isn't something keeping
> the socket open somehow ?
> 
> I don't know why a signal should cause this because the destruction of a
> socket should be the same code path whatever caused the process to exit.
> I wonder if you add a signal handler to the application which does:
> 
> 	close(socket);
> 	exit(0);
> 
> whether you still get the same problem ? What kernel version are you using ?
> 
> I know some of the DECnet code can be a bit slow in places in shutting down
> sockets as its overcautious, but I've not seen hangs in the RUN state
> before,

I've never seen it before but now it seems consistent, if I ^C a process with an
open connection it stays in RUN. I'm running kernel 2.4.17 

Oddly adding the signal handler makes no difference - the sockets still stick in
RUN (yes, I put a prinf in the signal handler so I knew it was being called!)

But they go away if I let the program run to completion.

Wierd.

patrick


P.S. Just noticed this too - I think It'd be nice if the link numbers in
/proc/net/decnet was in decimal rather than hex then it is possible to match
them up with those shown by VMS.

Re: [Linux-decnet-user] Closing DECNET connections

[Steven W. <st...@gw...>]

Hi,

[snip]
> 
> 
> Steve?
> 
> patrick
> 
> 
Hmm. Not sure about that. When the socket gets closed (along with all the
other fds of that process it should get shutdown. The release function has
the code in it to force that. Are you sure that there isn't something keeping
the socket open somehow ?

I don't know why a signal should cause this because the destruction of a
socket should be the same code path whatever caused the process to exit.
I wonder if you add a signal handler to the application which does:

	close(socket);
	exit(0);

whether you still get the same problem ? What kernel version are you using ?

I know some of the DECnet code can be a bit slow in places in shutting down
sockets as its overcautious, but I've not seen hangs in the RUN state
before,

Steve.

Re: [Linux-decnet-user] Closing DECNET connections

[Patrick C. <pa...@ty...>]

On Tue, Apr 16, 2002 at 02:40:44PM +0200, Lhota Robert wrote:
> > Hello,
> > 
> > I am using dnping (Linux-DECNET) for testing the availability of VAX
> > machines. The timeout is set for 3 seconds. Unfortunately, dnping
> > program does not implement timeout feature, so I have to kill the
> > child-process if it does not respond in the timeout period. The side
> > effect of this is that the DECNET stack on Linux does not close the
> > connection to the VAX machine. The number of connections (users) on
> > VAX machines is increasing and cosuming resources. Does anybody know
> > how to clear these connections on the Linux side? Can I just restart
> > the decnet stack (/etc/init.d/decnet restart) or is there some other
> > way? It looks like to me that the decnet stack somehow remembers the
> > connections even after service is restarted.

You're right. It's not just dnping but any DECnet connections. They don't seem
to close when the application is aborted using ^C

$ cat /proc/net/decnet
Local                                              Remote
  1.10/201A 0001:0013 0000:0001 2 LINUX              1.30/201B 0000:0001 0000:0012 2 25                RUN IMMED
  1.10/201B 0001:0013 0000:0001 2 LINUX              1.30/201C 0000:0001 0000:0012 2 25                RUN IMMED
  1.10/201C 0001:0013 0000:0001 2 LINUX              1.30/201D 0000:0001 0000:0012 2 25                RUN IMMED
  1.10/201D 0001:0013 0000:0001 2 LINUX              1.30/201E 0000:0001 0000:0012 2 25                RUN IMMED
  1.10/201E 0001:0006 0000:0001 2 PATRICK            1.30/201F 0000:0001 0000:0005 2 17                RUN IMMED
  1.10/201F 0001:0005 0000:0001 2 PATRICK            1.30/2020 0000:0001 0000:0004 2 17                RUN IMMED
  1.10/2020 0001:0005 0000:0001 2 PATRICK            1.30/2021 0000:0001 0000:0004 2 17                RUN IMMED
  1.10/2021 0001:0005 0000:0001 2 PATRICK            1.30/2023 0000:0001 0000:0004 2 17                RUN IMMED
  1.10/2022 0001:0005 0000:0001 2 PATRICK            1.30/2024 0000:0001 0000:0004 2 17                RUN IMMED

$ mc ncp sho kno links
 
 
Known Link Volatile Summary as of 16-APR-2002 14:06:56
 
   Link       Node           PID     Process     Remote link  Remote user
 
  8219    1.10 (ARTHUR)    000000BC  MIRROR_8219        8218  LINUX
  8220    1.10 (ARTHUR)    000000BD  MIRROR_8220        8219  LINUX
  8221    1.10 (ARTHUR)    000000BE  MIRROR_8221        8220  LINUX
  8222    1.10 (ARTHUR)    000000BF  MIRROR_8222        8221  LINUX
  8223    1.10 (ARTHUR)    000000C1  FAL_8223           8222  PATRICK
  8224    1.10 (ARTHUR)    000000C2  FAL_8224           8223  PATRICK
  8225    1.10 (ARTHUR)    000000C3  FAL_8225           8224  PATRICK
  8227    1.10 (ARTHUR)    000000C4  FAL_8227           8225  PATRICK
  8228    1.10 (ARTHUR)    000000C5  FAL_8228           8226  PATRICK
  8231    1.10 (ARTHUR)    00000102  FAL_8231           8229  PATRICK

As a temporary workaround you can remove the connections using the 
NCP DISCONECT LINK 
command but it looks quite nasty.


Steve?

patrick

[Linux-decnet-user] Closing DECNET connections

[Lhota R. <rl...@co...>]

> Hello,
>=20
> I am using dnping (Linux-DECNET) for testing the availability of VAX
> machines. The timeout is set for 3 seconds. Unfortunately, dnping
> program does not implement timeout feature, so I have to kill the
> child-process if it does not respond in the timeout period. The side
> effect of this is that the DECNET stack on Linux does not close the
> connection to the VAX machine. The number of connections (users) on
> VAX machines is increasing and cosuming resources. Does anybody know
> how to clear these connections on the Linux side? Can I just restart
> the decnet stack (/etc/init.d/decnet restart) or is there some other
> way? It looks like to me that the decnet stack somehow remembers the
> connections even after service is restarted.
>=20
> Thanks in advance for the answer,
> Robert Lhota
>=20

[Linux-decnet-user] Closing DECNET connections

[Lhota R. <rl...@co...>]

Hello,

I am using dnping (Linux-DECNET) for testing the availability of VAX
machines. The timeout is set for 3 seconds. Unfortunately, dnping
program does not implement timeout feature, so I have to kill the
child-process if it does not respond in the timeout period. The side
effect of this is that the DECNET stack on Linux does not close the
connection to the VAX machine. The number of connections (users) on VAX
machines is increasing and cosuming resources. Does anybody know how to
clear these connections on the Linux side? Can I just restart the decnet
stack (/etc/init.d/decnet restart) or is there some other way? It looks
like to me that the decnet stack somehow remembers the connections even
after service is restarted.

Thanks in advance for the answer,
Robert Lhota
=20

Re: [Linux-decnet-user] Debian tools and where to find them and also what device settings should be used for the interface

[Patrick C. <pa...@ty...>]

On Mon, Apr 01, 2002 at 05:14:39PM -0500, Gregg C Levine wrote:
> Hello from Gregg C Levine
> In an earlier e-mail Patrick mentioned the fact that he maintains the
> Debian collection for the Linux-DECnet tool set. And that if I were
> running it here, all I needed to was to issue these commands, "'#
> apt-get install dnet-progs' 
> answer a couple of questions and reboot(to set the MAC address).".Well
> regrettably I am not. But I am wondering where to find the tool set
> referenced in that command prompt line. I checked the ftp site for
> collection, and found something else labeled "apt", but that did not
> look right. Found the site for the alien concept(?), and downloaded the
> stuff that I know to be referencing Slackware, and ran from there to the
> Debian site. 

I'm nore sure the debian tools would be much use to you on a Slackware
system, you really need to be running a Debian system to use them.

I suspect it would be easier to recompile dnprogs yourself than try to
get apt and it's friends running on a "foreign" distribution. 


> Also, I would like to know what settings I would need to
> use for setting the /dev/lat* settings? You know what I mean, is it a
> character or block, which major, and which minor, and such like. TCPdump
> complains, and I need to manually set the interface, ifconfig, also
> complains, but does tell me that my Ethernet connection is still there.

The /dev/lat devices are simply symlinks to real devices, actually PTYs. You
don't need to do anything with them at all - in fact there is nothing useful you
could do with them. latcp deals with all the management of them.


patrick

RE: [Linux-decnet-user] Bug in dnprogs-2.17

[Gregg C L. <dr...@wo...>]

Hello again from Gregg C Levine
Close. I unzipped the file onto a MSDOS formatted hard disk partition,
and worked with it there. But it refused to work correctly. I also tried
it here. Same story different problem. Basically with me, it=92s the =
video
card that I use here, this one is an ATI Rage 128 card, and with the
other guy, it's a Trident TGUI9440. You=92d be surprised as to the
problems with X, on those cards. What was your message? No, seriously, I
want to compare notes on problems with ZipSlack.. Oh, and because of a
snafu with one of my other ongoing projects, I temporarily reverted to
an older version of Slackware, the 7.1, version. The 2.2.16 patch worked
perfectly. First try, if possible, will be with that setup, and then
with 2.4.5.=20
-------------------
Gregg C Levine dr...@wo...
------------------------------------------------------------
"The Force will be with you...Always." Obi-Wan Kenobi
"Use the Force, Luke."=A0 Obi-Wan Kenobi
(This company dedicates this E-Mail to General Obi-Wan Kenobi )
(This company dedicates this E-Mail to Master Yoda )



> -----Original Message-----
> From: lin...@li...
[mailto:linux-decnet-user-
> ad...@li...] On Behalf Of Patrick Caulfield
> Sent: Tuesday, April 02, 2002 2:39 AM
> To: lin...@li...
> Subject: Re: [Linux-decnet-user] Bug in dnprogs-2.17
>=20
> On Mon, Apr 01, 2002 at 02:50:44PM -0500, Gregg C Levine wrote:
> > Hello from Gregg C Levine
> > And here's the funny part. Last night, after posting that note, I
> > repeated my steps. It worked. It had a problem figuring out what I
am
> > running. I had added a few things found on Red Hat systems. Now it
> > works. But all I need now is a DEC system to connect to...... Oh,
and I
> > tried out, the ZipSlack version from the 8.00 repository, it did not
> > work on that guy. Mores the pity. Since the founders encourage
people to
> > send out those bug reports, I figure I'd do that.
>=20
> Unless it the same message I suspect that's more due to limitations on
what they
> can fit onto a 100Meg installation.
>=20
> patrick
>=20
>=20
> _______________________________________________
> Linux-decnet-user mailing list
> Lin...@li...
> https://lists.sourceforge.net/lists/listinfo/linux-decnet-user
>=20

Re: [Linux-decnet-user] Bug in dnprogs-2.17

[Patrick C. <pa...@ty...>]

On Mon, Apr 01, 2002 at 02:50:44PM -0500, Gregg C Levine wrote:
> Hello from Gregg C Levine
> And here's the funny part. Last night, after posting that note, I
> repeated my steps. It worked. It had a problem figuring out what I am
> running. I had added a few things found on Red Hat systems. Now it
> works. But all I need now is a DEC system to connect to...... Oh, and I
> tried out, the ZipSlack version from the 8.00 repository, it did not
> work on that guy. Mores the pity. Since the founders encourage people to
> send out those bug reports, I figure I'd do that.

Unless it the same message I suspect that's more due to limitations on what they
can fit onto a 100Meg installation.

patrick

[Linux-decnet-user] Debian tools and where to find them and also what device settings should be used for the interface

[Gregg C L. <dr...@wo...>]

Hello from Gregg C Levine
In an earlier e-mail Patrick mentioned the fact that he maintains the
Debian collection for the Linux-DECnet tool set. And that if I were
running it here, all I needed to was to issue these commands, "'#
apt-get install dnet-progs'=20
answer a couple of questions and reboot(to set the MAC address).".Well
regrettably I am not. But I am wondering where to find the tool set
referenced in that command prompt line. I checked the ftp site for
collection, and found something else labeled "apt", but that did not
look right. Found the site for the alien concept(?), and downloaded the
stuff that I know to be referencing Slackware, and ran from there to the
Debian site. Also, I would like to know what settings I would need to
use for setting the /dev/lat* settings? You know what I mean, is it a
character or block, which major, and which minor, and such like. TCPdump
complains, and I need to manually set the interface, ifconfig, also
complains, but does tell me that my Ethernet connection is still there.

-------------------
Gregg C Levine dr...@wo...
------------------------------------------------------------
"The Force will be with you...Always." Obi-Wan Kenobi
"Use the Force, Luke."=A0 Obi-Wan Kenobi
(This company dedicates this E-Mail to General Obi-Wan Kenobi )
(This company dedicates this E-Mail to Master Yoda )