Routing Problem
Rich Walker
I have very little unix knowledge. I'm trying to get a OS5.04Eb system to telnet reliably through a router. It works fine on the local area network. Through the router, if a command causes a "large" amount of data to be sent to the screen, then the output simply hangs. For example, I log on as root and type "ls -l" -- the only visible response is the CRLF, no other chars are sent to the screen. If I type "ls" it works quite reliably. When it hangs up, the system does not respond to any further keystrokes. However, it will disconnect and reconnect readily.
To me it seems like a buffering problem in that, apparently if the SCO machine tries to send a large output via the router something fails, but many small transers are fine. Perhaps there is a parameter that I need to set?? Some info follows:
# route get default
route to: default
destination: default
mask: default
gateway: 209.61.22.65
interface: net0
flags: <UP,GATEWAY,STATIC,PMTU>
proto: 2, age: 77036
rpipe spipe ssthresh rtt,msec rttvar hopcount mtu expire tos
0 0 0 0 0 0 1500 0 0
# netstat -i
Name Mtu Network Address Ipkts Ierrs Opkts Oerrs Coll
net0 1500 209.61.22 scosysv 6746 0 3616 2 9
lo0 8232 loopback localhost 33736 0 33736 0 0
atl0* 8232 none none No Statistics Available
The router is a "compatible systems" model 1250i connected to TIAC
(owned by PSInet) hub in hartford CT.
Thanks for any help
Rich Walker
Tony Earnshaw
> I have very little unix knowledge. I'm trying to get a OS5.04Eb system
> to telnet reliably through a router. It works fine on the local area
> network. Through the router, if a command causes a "large" amount of
> data to be sent to the screen, then the output simply hangs. For
> example, I log on as root and type "ls -l" -- the only visible
> response is the CRLF, no other chars are sent to the screen. If I type
> "ls" it works quite reliably. When it hangs up, the system does not
> respond to any further keystrokes. However, it will disconnect and
> reconnect readily.
> To me it seems like a buffering problem in that, apparently if the SCO
> machine tries to send a large output via the router something fails,
> but many small transers are fine. Perhaps there is a parameter that I
> need to set?? Some info follows:
I suppose my initial reaction was: "there are an awful lot of OS 5.0.x
machines connected to the Internet via routers and the vast majority of
them telnet perfectly well to machines in other domains. So why doesn't
this one?"
Then I thought: "pity he doesn't tell us what kind of topography he uses
to get to the router, what NIC he's using, driver version, configured
MTU, or recv size send size, patches etc". Because they've all probably
got to do with it. Especially the NIC, driver version and patches.
Tony
--
************* THE NEW DIMENSION IN DISTRIBUTION ***********
ilion Faculty B.V.
Tony Earnshaw email: to...@ilion.nl
Randstad 21-57
1314 BH Almere-Stad tel: +31 (0) 36 548 50 10
The Netherlands fax: +31 (0) 36 534 05 34
***************** http://www.ilion.nl *********************
Rich Walker
available as the server is at a remote location. However, I've
determined that the problem is repeatable and related to the size of a
single buffer transfer. I can successfully cat a text file of 1270
bytes which results in a buffer transfer of 1458 bytes from the router
to my telnet client. If I add a single char to the file, then the cat
fails every time with no buffer transfered at all.
How can I inspect and change the send size? Since the MTU setting is
1500, that looks suspiciously close the boundary that I'm seeing with my
testing. As shown in my first post, the MTU size is 1500 for net0
through the router, but 8232 for the loopback path. How can I increase
this number from 1500?
Thanks,
Rich
Rich Walker
# ifconfig net0
net0: flags=4043<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 209.61.22.94 netmask ffffff00 broadcast 209.61.22.255
perf. params: recv size: 24576; send size: 24576; full-size
frames: 1
ether 00:60:08:cd:c7:eb
Jeff Liebermann
wrote:
>I have very little unix knowledge. I'm trying to get a OS5.04Eb system to telnet reliably through a router. It works fine on the local area network. Through the router, if a command causes a "large" amount of data to be sent to the screen, then the output simply hangs.
1. Could you please format your posting to about 75 characters per line so
I don't have to do it for you? It's kinda messy and difficult to read.
2. Please lose the Vcard signature.
3. There's not enough info supplied to make a decent guess. So, you get a
wild guess. Try looking at these and see if they apply. As I vaguely
recall, telnet hangs were fixed with OSS499A for 3.2v5.0.4.
Telnet sessions hang when sending large amounts of data.
http://www.sco.com/cgi-bin/ssl_reference?107652
When using the ftp protocol to ftp large files, ftp hangs.
http://www.sco.com/cgi-bin/ssl_reference?107664
My telnet(TC), rlogin(TC) or ftp(TC) session hangs when using a Compaq
NetFlex-2 card.
http://www.sco.com/cgi-bin/ssl_reference?104982
OSS449 docs and download
http://www.sco.com/cgi-bin/ssl_getsupplement?oss449
See:
http://www.sco.com/ta/
for additional techy articles and bug fixes.
--
Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
(831)421-6491 pgr (831)426-1240 fax (831)336-2558 home
http://www.cruzio.com/~jeffl WB6SSY
je...@comix.santa-cruz.ca.us je...@cruzio.com
Jeff Liebermann
Cool. This is almost as good as a detective story.
It tells me that 00:60:08 is 3com 3C905TX card. See:
http://cs.okidata.com/mantr320.nsf/Contents?OpenView
for how I did that trick.
NS lookup tells me that 209.61.22.94 is pingable and sitting on tiac.net.
Reverse DNS says nobody listed. Traceroute shows that it's in Hartford
Connecticut. I can ping it so I guess it's working. Oh, you already said
that in your original posting. You really should not post customers IP's
and system names.
whois at ARIN for the net block shows:
Financial Datacorp (NETBLK-TIAC-FDATA) TIAC-FDATA
209.61.22.64 - 209.61.22.95
I don't think they'll be too happy with either of us if I try to break in.
However, I'm bored and need some entertainment.
Firing up SNMP shows that your system name is scosysv.UUCP.com which is
rediculous and should be fixed. See:
http://www.cruzio.com/~jeffl/sco/new_name.txt
Ummm... No firewall?. You really should get a firewall instead of just a
router. I'm partial to Sonicwall/10 this week:
http://www.sonicsys.com
It also shows that you're running:
SCO TCP/IP Runtime Release 2.0.0
which means that this is 3.2v5.0.x.
Your system has been up for 3 days 40 hrs and 10 minutes.
Your 3c9905TX ethernet card is running at 10mbits/sec. This if this is NOT
correct, you may want to hack the space.c file for your ethernet card to
disarm the possibly broken autodetect feature and force the correct speed
and mode.
You've moved 4.2GBytes of data in the last 3 days on this machine so I can
fairly well presume that it's working.
Since your broadcast address is currently 209.61.22.255, the router might
not see the broadcasts. Judging by your number assignments, the correct
broadcast address should be 209.61.22.95 although I think .255 might work.
Whatever is at at 209.61.22.95 returns a MAC address of 0000A5E09700 which
is owned by Compatible Systems Corporation. So, what did PSI.net say to
use for a broadcast address and why is there a router sitting where I would
expect the broadcast address to be? Normally (if there is such a thing,
the router address is the lowest number of the IP block, while the
broadcast address is the highest.
Your ISP's gateway is at 209.61.22.65 with only a little traffic. Is this
the "router"? I don't see anything that even looks like a router exept for
the one at your ISP's at .65. Let me guess... Someone added a router or
machine and used the broadcast address for the IP. That won't work.
SNMP shows that your OSR5 box is receiving route metric values from
somewhere, probably the router. Please disarm, kill, and never resurrect
RIP from the OSR5 box. Edit /etc/tcp, find the lines that start "routed"
and put a comment symbol # in front of the routed line. Similarly, kill
the routed daemon if it's running.
The connection table shows that your OSR5 box is listening on a
considerable number of IP ports. You *REALLY* should get a firewall or at
least a packet filter. TCP Wrappers? Anything?
OK, enough entertainment value for one day. Back to work...
Jeff Liebermann
<je...@comix.santa-cruz.ca.us> wrote:
>Cool. This is almost as good as a detective story.
>It tells me that 00:60:08 is 3com 3C905TX card. See:
Oops. Wrong URL. Try:
http://www.cavebear.com/CaveBear/Ethernet/vendor.html
for the list of MAC prefixes.
Eric Trudel
coming from a Novell Netware 5 File Server:
Jan 17 11:04:32 hostname routed [333]: packet from unknown router,
172.21.45.11
I dont have those broadcasts into an AIX server.
Is someone can help me on that ?
Thanks.
Matthew Schalit
|># ifconfig net0
|>net0: flags=4043<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
|> inet 209.61.22.94 netmask ffffff00 broadcast 209.61.22.255
|> perf. params: recv size: 24576; send size: 24576; full-size
|>frames: 1
|> ether 00:60:08:cd:c7:eb
|Cool. This is almost as good as a detective story.
|...
|Since your broadcast address is currently 209.61.22.255, the router might
|not see the broadcasts. Judging by your number assignments, the correct
|broadcast address should be 209.61.22.95 although I think .255 might work.
|...
I'm not sure that I see your logic with the addressing, Jeff.
Did you get a different netmask from your probe besides what is listed
in ifconfig net0? If not, then, given that IP address and mask,
it looks like the bcast is correct.
209 . 61 . 22 . 94 IP Addy in Decimal
11010001 . 00111101 . 00010110 . 01011110 IP Addy in Binary
--------------------------------------------------------------------------
255 . 255 . 255 . 0 Netmask
11111111 . 11111111 . 11111111 . 00000000 Netmask in Binary
--------------------------------------------------------------------------
11010001 . 00111101 . 00010110 . 11111111 Bcast in Binary
209 . 61 . 22 . 255 Bcast in Decimal
So I'm curious what you're using for the netmask.
Take it easy,
Matt