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

Re: [Linux-decnet-user] Printing via Reverse LAT

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

On Tue, Dec 30, 2003 at 11:43:54PM +0100, Jens Wulf wrote:
> 
> ----- Original Message ----- 
> From: "Patrick Caulfield" <pa...@ty...>
> To: "Decnet" <lin...@li...>
> Sent: Sunday, December 28, 2003 3:57 PM
> Subject: Re: [Linux-decnet-user] Printing via Reverse LAT
> 
> 
> > One thing you could try to test my hypothesis is to put a delay before the
> > closing of the pipe, thus :
> >
> > (ls ge*; sleep 1) > /dev/lat/TS703
> >
> 
> Yes - this works!
> Isn't it possible to get latd to send all it gets after that delay ?
> 

I've put a possible fix on the head of CVS. It seems to work for me but I'd be
grateful if you could test it. 

There are other things in that part of CVS that may cause trouble (though I
hope they won't) as it's a bit of an experimental area at the moment but any
feedback would be appreciated.


patrick

Re: [Linux-decnet-user] Printing via Reverse LAT

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

On Tue, Dec 30, 2003 at 11:43:54PM +0100, Jens Wulf wrote:
> 
> ----- Original Message ----- 
> From: "Patrick Caulfield" <pa...@ty...>
> To: "Decnet" <lin...@li...>
> Sent: Sunday, December 28, 2003 3:57 PM
> Subject: Re: [Linux-decnet-user] Printing via Reverse LAT
> 
> 
> > One thing you could try to test my hypothesis is to put a delay before the
> > closing of the pipe, thus :
> >
> > (ls ge*; sleep 1) > /dev/lat/TS703
> >
> 
> Yes - this works!
> Isn't it possible to get latd to send all it gets after that delay ?
> 

Well, at least that confirms what I suspected. What actually seems to be
happening is that DATA messages are queued until the next circuit timer whereas
EOF messages are sent immediately. This is obviously a bug as, you have seen, if
the EOF happens when there is data pending then that data never gets sent.

So, as I mentioned, I need to delay the deletion of circuits untill all the data
has been transmitted. I'll look into it.


patrick

Re: [Linux-decnet-user] Printing via Reverse LAT

[<jen...@t-...>]

----- Original Message ----- 
From: "Patrick Caulfield" <pa...@ty...>
To: "Decnet" <lin...@li...>
Sent: Sunday, December 28, 2003 3:57 PM
Subject: Re: [Linux-decnet-user] Printing via Reverse LAT


> One thing you could try to test my hypothesis is to put a delay before the
> closing of the pipe, thus :
>
> (ls ge*; sleep 1) > /dev/lat/TS703
>

Yes - this works!
Isn't it possible to get latd to send all it gets after that delay ?


Thanks so far

Jens

Re: [Linux-decnet-user] Printing via Reverse LAT

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

One thing you could try to test my hypothesis is to put a delay before the
closing of the pipe, thus :

(ls ge*; sleep 1) > /dev/lat/TS703


patrick

[Linux-decnet-user] Printing via Reverse LAT

[<jen...@t-...>]

Just another information :

I recorded the communication between the Linux-machine and the DS 700 with
ethereal,

when using "dir ge*" then there s much communication, but the wanted Data is
not transmitted to the server.

Greets

Jens

Re: [Linux-decnet-user] Printing via Reverse LAT (latd Version 1.19)

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

On Sun, Dec 28, 2003 at 12:02:21PM +0100, Jens Wulf wrote:
> Hi,
>
> in addition to my last post i tried to debug this problem, but my Knowledge
> about C is too poor.

Hmmm it looks like a timing problem. In the "ge*" case the EOF arrives very
quickly and gets sent before the next circuit timer. At a guess (and it is only
a guess) it arrives at the terminal server before it has had a chance to flush
it's buffers of the data it just got and so the data gets thrown away.

Maybe what ought to happen is that EOF disconnects should be sent on the circuit
timer as the data messages are ... that causes a problem with the current code
as the comment says. The way to fix this is to queue sessions and connections
for deletion and remove them on the circuit timer. I'll look into this but it's
not a trivial fix and I'm desperately short of time too.

patrick

[Linux-decnet-user] Printing via Reverse LAT (latd Version 1.19)

[<jen...@t-...>]

Hi,

in addition to my last post i tried to debug this problem, but my Knowledge
about C is too poor.

Anyway i added a debug-line in File session.cc as follows :

....
int LATSession::read_pty()
...
    msglen = read(master_fd, buf, max_read_size);
    // ------- for testing
    debuglog(("**** Session : %d From PTY(%d): '%s%s'\n", local_session,
msglen,
    buf, (msglen>10)?"...":""));
...

and got the following debug-logs

Debug-log of the working - connection (Command : "dir g* >/dev/lat/TS703")

   32: LocalPort::do_read(), connected: 0
   32: localport::connect_session: master-fd = 6
   32: Looking for connection to node TS703
   32: New client connection for TS703 created
   32: localport for TS703 has connid 1
   32: new client session: localid 1, remote id 0
   32: new llogin session: localid 1, remote id 0
   32: new localport session: localid 1, remote id 0, fd=6
   32: lloginSession::new_session
   32: connecting parent for /dev/lat/TS703
   32: Requesting connect to service on interface 3
   32: Sending message for connid 1 (seq: 0, ack: 255, msg: 6) window=1
   32: set_fd_state: 6, 1
   32: LocalPort:: do_read disabling pty reads
   32: set_fd_state: 6, 1
   32: Got connect ACK for 1
   32: got connect ack. seq: 0, ack: 0
   32: connecting llogin session to 'TS703'
   32: Sending message for connid 1 (seq: 1, ack: 0, msg: 2) window=1
   32: session cmd for connid 1
   32: process_session_cmd: 2 slots, 64 bytes
   32: process_slot_cmd(0:8). command: 90, credit: 2, len: 24
   32: Session:: got connection for rem session 1
   32: set_fd_state: 6, 0
   32: Queued messsge for connid 1
   32: process_slot_cmd(1:24). command: a0, credit: 0, len: 23
   32: Sending 1 slots in reply
   32: Sending message for connid 1 (seq: 2, ack: 1, msg: 2) window=0
   32: LocalPort::do_read(), connected: 1
   32: do_read() called for LocalPort on connected socket
   32: set_fd_state: 6, 1
   32: session cmd for connid 1
   32: process_session_cmd: 0 slots, 46 bytes
   32: No data: cmd: 0, credit: 0
   32: service announcement: status = 2
   32: Got service. Node: TS703, service TS703, rating: 90
   32: **** Session : 1 From PTY(250): '-rw-r--r--    1 root     root
6836 Dec 11 22:07 geht
-rw-------    1 root     root          160 Dec 11 21:59 gehtnichtmQ
-rw-------    1 root     root          236 Dec 11 21:50 gehtnichtoQ
-rw-r--r--    1 ro(some unprintable characters here)...'
   32: Sending more remote credit
   32: set_fd_state: 6, 1
   32: Out of credit...Stop
   32: circuit Timer:: slots pending = 1
   32: Collected 1 slots on circuit timer
   32: Pending messages. Window size: 0, max 1. last_ack=2, last_seq=2
   32: Sending message on circuit timer: seq: 3, ack: 2
   32: session cmd for connid 1
   32: process_session_cmd: 1 slots, 46 bytes
   32: process_slot_cmd(0:8). command: a0, credit: 1, len: 23
   32: Got some more credit, (+1=1) carrying on
   32: set_fd_state: 6, 0
   32: Sending 1 slots in reply
   32: Sending message for connid 1 (seq: 4, ack: 3, msg: 2) window=0
   32: LocalPort::do_read(), connected: 1
   32: do_read() called for LocalPort on connected socket
   32: set_fd_state: 6, 1
   32: session cmd for connid 1
   32: process_session_cmd: 0 slots, 46 bytes
   32: No data: cmd: 0, credit: 0
   32: **** Session : 1 From PTY(60): 'ot     root        10028 Dec  4 19:50
gmon.out
...'
   32: set_fd_state: 6, 1
   32: Out of credit...Stop
   32: circuit Timer:: slots pending = 1
   32: Collected 1 slots on circuit timer
   32: Pending messages. Window size: 0, max 1. last_ack=4, last_seq=4
   32: Sending message on circuit timer: seq: 5, ack: 4
   32: session cmd for connid 1
   32: process_session_cmd: 0 slots, 46 bytes
   32: No data: cmd: 0, credit: 0
   32: Pending messages. Window size: 0, max 1. last_ack=5, last_seq=5
   32: Sending message on circuit timer: seq: 6, ack: 5
   32: session cmd for connid 1
   32: process_session_cmd: 0 slots, 46 bytes
   32: No data: cmd: 0, credit: 0
   32: session cmd for connid 1
   32: process_session_cmd: 1 slots, 46 bytes
   32: process_slot_cmd(0:8). command: 0, credit: 1, len: 0
   32: Got some more credit, (+1=1) carrying on
   32: set_fd_state: 6, 0
   32: 0 echo bytes available
   32: Remote credit is 5
   32: Sending 0 slots in reply
   32: Sending message for connid 1 (seq: 7, ack: 7, msg: 2) window=0
   32: LocalPort::do_read(), connected: 1
   32: do_read() called for LocalPort on connected socket
   32: set_fd_state: 6, 1
   32: session cmd for connid 1
   32: process_session_cmd: 0 slots, 46 bytes
   32: No data: cmd: 0, credit: 0
   32: session cmd for connid 1
   32: process_session_cmd: 1 slots, 46 bytes
   32: process_slot_cmd(0:8). command: 0, credit: 1, len: 0
   32: 0 echo bytes available
   32: Remote credit is 5
   32: Sending 0 slots in reply
   32: Sending message for connid 1 (seq: 8, ack: 9, msg: 2) window=0
   32: session cmd for connid 1
   32: process_session_cmd: 0 slots, 46 bytes
   32: No data: cmd: 0, credit: 0
   32: **** Session : 1 From PTY(-1): ''
   32: EOF on PTY
   32: Sending message for connid 1 (seq: 9, ack: 10, msg: 2) window=1
   32: lloginSession::disconnect_session()
   32: set_fd_state: 6, 0
   32: LocalPort::do_read(), connected: 1
   32: do_read() called for LocalPort on connected socket
   32: set_fd_state: 6, 1
   32: remove_fd: 6
   32: Deleting session 1
   32: remove_fd: -1
   32: Restarting PTY for local session
   32: LocalPort::restart_pty()
   32: set_fd_state: 6, 1
   32: remove_fd: 6
   32: openpty: master_fd=6, slave_fd=7
   32: made symlink /dev/lat/TS703 to /dev/pts/3
   32: remove_fd: -1
   32: connection 1 pending deletion
   32: LATConnection dtor: 1
   32: session cmd for connid 1
   32: service announcement: status = 2
   32: Got service. Node: TS703, service TS703, rating: 255

 ---------------------------------------------------------------------------
----

 Debug-Log of not-working Connection (Command : dir ge* >/dev/lat/TS703)

  146: LocalPort::do_read(), connected: 0
  146: localport::connect_session: master-fd = 6
  146: Looking for connection to node TS703
  146: New client connection for TS703 created
  146: localport for TS703 has connid 2
  146: new client session: localid 1, remote id 0
  146: new llogin session: localid 1, remote id 0
  146: new localport session: localid 1, remote id 0, fd=6
  146: lloginSession::new_session
  146: connecting parent for /dev/lat/TS703
  146: Requesting connect to service on interface 3
  146: Sending message for connid 2 (seq: 0, ack: 255, msg: 6) window=1
  146: set_fd_state: 6, 1
  146: LocalPort:: do_read disabling pty reads
  146: set_fd_state: 6, 1
  146: Got connect ACK for 2
  146: got connect ack. seq: 0, ack: 0
  146: connecting llogin session to 'TS703'
  146: Sending message for connid 2 (seq: 1, ack: 0, msg: 2) window=1
  146: session cmd for connid 2
  146: process_session_cmd: 2 slots, 64 bytes
  146: process_slot_cmd(0:8). command: 90, credit: 2, len: 24
  146: Session:: got connection for rem session 1
  146: set_fd_state: 6, 0
  146: Queued messsge for connid 2
  146: process_slot_cmd(1:24). command: a0, credit: 0, len: 23
  146: Sending 1 slots in reply
  146: Sending message for connid 2 (seq: 2, ack: 1, msg: 2) window=0
  146: LocalPort::do_read(), connected: 1
  146: do_read() called for LocalPort on connected socket
  146: set_fd_state: 6, 1
  146: session cmd for connid 2
  146: process_session_cmd: 0 slots, 46 bytes
  146: No data: cmd: 0, credit: 0
  146: service announcement: status = 2
  146: Got service. Node: TS703, service TS703, rating: 90
  146: **** Session : 1 From PTY(235): '-rw-r--r--    1 root     root
6836 Dec 11 22:07 geht
-rw-------    1 root     root          160 Dec 11 21:59 gehtnichtmQ
-rw-------    1 root     root          236 Dec 11 21:50 gehtnichtoQ
(some unprintable characters here)...'
  146: Sending more remote credit
  146: set_fd_state: 6, 1
  146: Out of credit...Stop
  146: circuit Timer:: slots pending = 1
  146: Collected 1 slots on circuit timer
  146: Pending messages. Window size: 0, max 1. last_ack=2, last_seq=2
  146: Sending message on circuit timer: seq: 3, ack: 2
  146: session cmd for connid 2
  146: process_session_cmd: 1 slots, 46 bytes
  146: process_slot_cmd(0:8). command: a0, credit: 1, len: 23
  146: Got some more credit, (+1=1) carrying on
  146: set_fd_state: 6, 0
  146: Sending 1 slots in reply
  146: Sending message for connid 2 (seq: 4, ack: 3, msg: 2) window=0
  146: LocalPort::do_read(), connected: 1
  146: do_read() called for LocalPort on connected socket
  146: set_fd_state: 6, 1
  146: session cmd for connid 2
  146: process_session_cmd: 0 slots, 46 bytes
  146: No data: cmd: 0, credit: 0
  146: **** Session : 1 From PTY(-1): ''
  146: EOF on PTY
  146: Sending message for connid 2 (seq: 5, ack: 4, msg: 2) window=1
  146: lloginSession::disconnect_session()
  146: set_fd_state: 6, 0
  146: Pending messages. Window size: 1, max 1. last_ack=4, last_seq=5
  146: LocalPort::do_read(), connected: 1
  146: do_read() called for LocalPort on connected socket
  146: set_fd_state: 6, 1
  146: remove_fd: 6
  146: Deleting session 1
  146: remove_fd: -1
  146: Restarting PTY for local session
  146: LocalPort::restart_pty()
  146: set_fd_state: 6, 1
  146: remove_fd: 6
  146: openpty: master_fd=6, slave_fd=7
  146: made symlink /dev/lat/TS703 to /dev/pts/3
  146: remove_fd: -1
  146: connection 2 pending deletion
  146: LATConnection dtor: 2
  146: session cmd for connid 2
  146: service announcement: status = 2
  146: Got service. Node: TS703, service TS703, rating: 255
----------------------------------------------------------------------------
--------------------------

The first example prints on the port as it should, the second produces no
output.
I stuck here - hope anyone can help.

Greets

Jens

[Linux-decnet-user] latd 1.19 released

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

This is a minor update off the 1.18 branch in CVS. It mainly tidies up things I
should have done ages ago. Principally *BSD support. In fact all of this stuff
has been in CVS for months.

The changelog reads like this:

  * Some Darwin bpf interface fixes.
  * fix lat group on /dev/lat/ devices that was broken by my autoconf changes
    in 1.17
  * Fix memory leak caused by latcp -d & llogin -d
    (thanks to Sergei Gorbatov)

So I doubt it will affect many people but it's out there now at least. The more
radical changes I have in the head of CVS (plus some unmerged stuff) may go out
later but it'll probably be marked as unstable. Of course, if I was actually any
good at releases they would have easily distinguishable numbers but that's
unlikely to happen.

The source tarball and RPM (built on Fedora Core 1) are available from the
usualy sourceforge location. Here's a link:

http://sourceforge.net/project/showfiles.php?group_id=4993&package_id=6007&release_id=205087

Not much of a Chrismas present, but have a happy one anyway if you're that way
inclined.
-- 

patrick

[Linux-decnet-user] Printing via Reverse LAT

[<li...@t-...>]

Hi
i tried to print to a DECSERVER 700 and have the following Problem :

using the following String in /etc/latd.conf :

$LATCP -A -p /de/lat/TS703 -V TS703 -R PORT_15 -Q

the i did : 

dir >/dev/lat/TS703

and it prints the directory as it should

when i do :

dir g* >/dev/lat/TS703     (2 files are printed on the screen with dir g*)

nothing is printed.

i don t know why - can anyone help ?

Greetings Jens

[Linux-decnet-user] State of DECnet in 2.6

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

This is just a test message really but I thought I'd put some (slightly) useful
content in it to make it less annoying :-)

DECnet in Linux 2.6 works fine apart from three known problems:

1. SDF_WILD sockets don't work. (ie dnetd won't start) The patch below fixes
this:

--- net/decnet/af_decnet.c.orig 2003-12-08 11:27:59.000000000 +0000
+++ net/decnet/af_decnet.c      2003-12-08 11:28:34.000000000 +0000
@@ -163,7 +163,7 @@
        struct dn_scp *scp = DN_SK(sk);

        if (scp->addr.sdn_flags & SDF_WILD)
-               return hlist_empty(&dn_wild_sk) ? NULL : &dn_wild_sk;
+               return hlist_empty(&dn_wild_sk) ? &dn_wild_sk : NULL;

        return &dn_sk_hash[scp->addrloc & DN_SK_HASH_MASK];
 }


2. cat /proc/net/decnet_cache will crash the machine. This patch fixes that
problem:

--- net/decnet/dn_route.c  24 Nov 2003 03:16:28 -0000      1.3
+++ net/decnet/dn_route.c  17 Dec 2003 09:54:37 -0000
@@ -1720,7 +1720,8 @@

 static void dn_rt_cache_seq_stop(struct seq_file *seq, void *v)
 {
-       rcu_read_unlock();
+       if (v)
+               rcu_read_unlock();
 }
  
 static int dn_rt_cache_seq_show(struct seq_file *seq, void *v)



3. Big-endian machines are broke.

I'm looking into the last one and will, with luck, get something done over
Christmas. With the two patches it should be fully functional on Intel and Alpha
platforms. 

Patrick

[Linux-decnet-user] Reverse Lat with DS700

[<jen...@t-...>]

Hi,
i tried to get up a printer-port with a Dec Server 700.

i tried several settings :

1. latcp -A -p /VTS701 /HTS701/RPORT_16 -Q
2. latcp -A -p /HTS701 /RPORT_16 -Q
3. latcp -A -p /HTS701 /RPORT_16

but i did't get anyone of them to work properly -
only the first variant worked one time - but i couldn t reproduce it.
printing for a Alpha with VMS works .
the 2nd and 3d variants only produced broadcast ( message to 09-00-2b ...)
but nothing else happens.
the port is configured with (Access : Remote and Type : Hard)
for the first variant i processed a DEFINE SERVICE TS701 Port 16.
Has anyone better success with a DS 700 ?

Greetings

Jens