Linux-Networking Digest #589, Volume #12 Tue, 14 Sep 99 17:13:35 EDT
Contents:
Samba 2.0.5a: session setup rejected by PDC and BDCs (NT4 SP5) (Martin Emmerich)
Re: Apache, ASP, and ODBC (Stuart Children)
Re: modem is hanging during pppd ([EMAIL PROTECTED])
Re: Can't get Sendmail to relay in RH6 (blackrose)
Re: ftp and masquerading (Vilmos Soti)
Re: named fails since upgrading from RH5.1 to RH5.2 (Mark Worsdall)
pppd stays up will not disconnect after idle time reached (Mark Worsdall)
Can't get Sendmail to relay in RH6 (blackrose)
Re: Routing problems ("Justin")
Re: Why does hitting a key seem to wake up TCP/IP? (David C.)
Re: TCP/IP Programming Mystery (David C.)
One more question ("Blacka")
SATAN ("Cjk Didelot")
MSCHAP81 ??? ("Russ W. Knize")
Re: Good Book on Firewalling - IPChains ? ("Edward Lee Ah Yen (AKA SMASH)")
firewalling a larger network ("Justin")
Port 106 (Kevin Wall)
Re: Sub-C networks? (David C.)
Re: Sub-C networks? (David C.)
Re: Recommendation for 100Mbps Switched Ethernet hardware (David C.)
Help! Inetd doesn't respond to requests. ("Jerry O'Brien")
Re: Problem to reach 100Mbps with 3C905B-TX (Maarten Afman)
----------------------------------------------------------------------------
Crossposted-To: comp.protocols.smb
From: [EMAIL PROTECTED] (Martin Emmerich)
Subject: Samba 2.0.5a: session setup rejected by PDC and BDCs (NT4 SP5)
Date: Tue, 14 Sep 1999 14:38:14 GMT
Hi,
I4ve setup a Samba 2.0.5a on a SuSe 6.2 based Linux 2.2.10. I4m
trying to integrate the server into our NT domain SF-DOM as
described in samba/textdocs/DOMAIN_MEMBER.txt . The described
procedures work, but access as domain user is impossible. In the
logfile I get:
[1999/09/14 11:18:31, 2] smbd/server.c:main(716)
Changed root to /
[1999/09/14 11:18:31, 2] smbd/reply.c:reply_special(96)
netbios connect: name1=SFM-LINUX1 name2=PLAN9
[1999/09/14 11:18:31, 0]
smbd/password.c:domain_client_validate(1276)
domain_client_validate: machine 10.1.0.10 rejected the session
setup. Error was : code 131.
[1999/09/14 11:18:31, 0]
smbd/password.c:domain_client_validate(1276)
domain_client_validate: machine 10.1.10.2 rejected the session
setup. Error was : code 131.
[1999/09/14 11:18:32, 0]
smbd/password.c:domain_client_validate(1276)
domain_client_validate: machine 10.1.10.5 rejected the session
setup. Error was : code 131.
[1999/09/14 11:18:32, 0]
smbd/password.c:domain_client_validate(1330)
domain_client_validate: Domain password server not available.
[1999/09/14 11:18:32, 1] smbd/password.c:pass_check_smb(504)
Couldn't find user 'memmerich' in smb_passwd file.
[1999/09/14 11:18:32, 2] smbd/reply.c:reply_sesssetup_and_X(830)
NT Password did not match for user 'memmerich' ! Defaulting to
Lanman
[1999/09/14 11:18:32, 1] smbd/password.c:pass_check_smb(504)
Couldn't find user 'memmerich' in smb_passwd file.
[1999/09/14 11:18:32, 2] smbd/server.c:exit_server(406)
Closing connections
The IP addresses of PDC and BDCs, server (SFM-LINUX1),
workstation (PLAN) and user name (memmerich) are correct.
The samba config file is:
# Global parameters
[global]
workgroup = SF-DOM
server string = Linux 2.2.10 (SuSe 6.2) Fileserver
security = domain
encrypt passwords = yes
map to guest = Bad User
password server = 10.1.0.10, 10.1.10.2, 10.1.10.5
log level = 2
syslog = 2
keepalive = 30
local master = no
wins server = 10.1.0.10
remote announce = 10.1.255.255
printing = bsd
load printers = yes
character set = iso8859-1
[share1]
comment = Share 1
path = /usr/share1
read only = no
browseable = yes
# guest ok = yes
# force user = nobody
# force group = nogroup
# create mask = 0777
# directory mask = 0777
[share2]
comment = Share 2
path = /usr2/share2
read only = no
browseable = yes
# guest ok = yes
# force user = nobody
# force group = nogroup
# create mask = 0777
# directory mask = 0777
If I uncomment the guest logins and set security to server, I
can anonymously browse the shares and create files there.
Whats wrong in the security=domain mode?
--
Martin Emmerich
[EMAIL PROTECTED]
"Wherever you go, there you are." (Buckaroo Banzai)
------------------------------
From: [EMAIL PROTECTED] (Stuart Children)
Subject: Re: Apache, ASP, and ODBC
Date: Tue, 14 Sep 1999 20:25 +0100 (BST)
Reply-To: [EMAIL PROTECTED]
Tim Bishop ([EMAIL PROTECTED]) wrote:
> I'm currently running Microsoft IIS4 using ASP and ODBC to use an Access
> database.
Yuk. ;)
> Is it possible to use ASP on Apache ? If not, then I suppose there is a
> 'similar' type of thing that I can use, but I don't know what I'm
> looking for !
OK. What I'd do would be to install PHP (http://www.php.net). It's a
server-side scripting language, in the same fashion as ASP is. It can be
compiled with Apache, or used as a CGI. If you don't mind taking the
time, then learn it. It's nice. Of particular interest to you, it has
functions for loads of different databases, including ODBC which you could
use to talk to Access.
However, you can stick with your ASP. Look at
http://www.nodeworks.com/asp/. It allows you to serve ASP pages on an
Apache server with mod_perl (http://perl.apache.org/) installed. I
haven't used it myself, although I'd be interested to hear from anyone
that has, as a friend of mine wanted me to host some ASP pages (he doesn't
want to learn PHP now he knows ASP) on my (Linux) machine.
There's also a program out there that will convert ASP scripts to PHP
ones. I don't have a URL for it I'm afraid. A look at the PHP homepage
might reveal something.
HTH,
- Stuart -
------------------------------
From: [EMAIL PROTECTED]
Crossposted-To: comp.os.linux.setup
Subject: Re: modem is hanging during pppd
Date: Tue, 14 Sep 1999 19:23:19 GMT
In article <7r9l1g$k4a$[EMAIL PROTECTED]>,
[EMAIL PROTECTED] (Bill Unruh) wrote:
> In <[EMAIL PROTECTED]> Leolo
<[EMAIL PROTECTED]> writes:
>
> >When i transmit ++ + (w/o the space, eh) over a
> >ppp connection, the
>
> Your modem has problems. There is guard time, by
> default 1 sec, which is supposed to be between
> the last character sent before and the first +
> and the last + and the first charactere sent
> afterwards. That guard time means that almost
> never does this happen during regular running.
> It sounds to me like you have somehow set that
> guard time to zero, thinking you were doing
> something else.
I just looked in
/etc/sysconfig/network-scripts/chat-ppp2 and it
had S12=0! S12 is the guard time in milliseconds.
So this was effectively removing the guard time.
Now, what was I on when I set that?
Thanks to the you all who pointed me in the right
direction.
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
------------------------------
From: [EMAIL PROTECTED] (blackrose)
Subject: Re: Can't get Sendmail to relay in RH6
Date: Tue, 14 Sep 1999 19:30:24 GMT
On Tue, 14 Sep 1999 19:28:44 GMT, [EMAIL PROTECTED] (blackrose) said:
>I've tried using access, and doing a "make access."
Er, "make access.db."
NO DICE!
br
@-}-
------------------------------
From: Vilmos Soti <[EMAIL PROTECTED]>
Subject: Re: ftp and masquerading
Date: Tue, 14 Sep 1999 15:09:01 GMT
Martin Stockhammer wrote:
>
> Hi,
>
> I have a problem with ftp. I have a linux router for the local net.
> The router does masquerade and uses ipchains for firewalling.
>
> Now I want to ftp from my Windows-Client to the webserver.
> Getting files is no problem, but if I want to put files to the server.
> The ftp-client says the file is sent, but then hangs.
> And on the webserver there is a file with size 0.
>
> If I put the file from the router with ncftp there is no problem.
Hi,
Did you insmod the ip_masq_ftp.o module on the machine which performs
masquerading?
Vilmos
------------------------------
From: Mark Worsdall <[EMAIL PROTECTED]>
Subject: Re: named fails since upgrading from RH5.1 to RH5.2
Date: Tue, 14 Sep 1999 19:33:03 +0100
In article <[EMAIL PROTECTED]>, Cowles, Steve
<[EMAIL PROTECTED]> writes
>Mark,
>
>The reason named-bootconf.pl just "sits there" is because you probably did not
>specify the
>input file, like /etc/named.boot!!!! The program is waiting for input... which
>typically
>is specified as a command line argument (like /etc/named.boot) or comes in from
>a pipe.
>
>try one of the following commands, they all worked on my system (RH60)
>
>named-bootconf.pl /etc/named.boot >/etc/named.conf
>
>or
>
>named-bootconf.pl </etc/named.boot >/etc/named.conf
>
>or even...
>
>cat /etc/named.boot | named-bootconf.pl >/etc/named.conf
>
Perfect, thanks, this thread can now be dead:-)
--
Mark Worsdall - Oh no, I've run out of underpants :(
Home:- [EMAIL PROTECTED] http://www.wizdom.org.uk
Shadow:- [EMAIL PROTECTED] http://www.shadow.org.uk
Work:- [EMAIL PROTECTED] http://www.hinwick.demon.co.uk
Web site Monitoring:- http://www.shadow.org.uk/SiteSight/
------------------------------
From: Mark Worsdall <[EMAIL PROTECTED]>
Subject: pppd stays up will not disconnect after idle time reached
Date: Tue, 14 Sep 1999 20:22:09 +0100
Hi,
Cannot get pppd to obey the ifcfg-ppp0 options, any ideas what is wrong?
I just want it to disconnect from the net after an idle time of 60
seconds but connect if squid demands a connecton:-
DEVICE="ppp0"
ONBOOT=no
USERCTL=yes
MODEMPORT=/dev/modem
LINESPEED=115200
NOPERSIST=no
DEMAND=yes
IDLE=30
DEFABORT=yes
DEBUG=yes
INITSTRING=ATZ
DEFROUTE=yes
HARDFLOWCTL=yes
ESCAPECHARS=no
PPPOPTIONS=
PAPNAME=
REMIP=
NETMASK="255.255.0.0"
IPADDR=
MRU=
MTU=
DISCONNECTTIMEOUT=60
RETRYTIMEOUT=5
BOOTPROTO=none
--
Mark Worsdall - Oh no, I've run out of underpants :(
Home:- [EMAIL PROTECTED] http://www.wizdom.org.uk
Shadow:- [EMAIL PROTECTED] http://www.shadow.org.uk
Work:- [EMAIL PROTECTED] http://www.hinwick.demon.co.uk
Web site Monitoring:- http://www.shadow.org.uk/SiteSight/
------------------------------
From: [EMAIL PROTECTED] (blackrose)
Subject: Can't get Sendmail to relay in RH6
Date: Tue, 14 Sep 1999 19:28:44 GMT
I don't care what ANYBODY says, I can't make relaying work in 6. 5.2
was a breeze. This is a fucking nightmare.
I've tried using access, and doing a "make access." No dice. I've
tried using "relay-domains" - no dice. I've tried adding the
F{LocalIP} rule in sendmail and using ip_allow. No dice. Nothing I do
will make my RH6 box relay mail for me from my local network
(192.168.1.0/24).
I tried reading the sendmail readme. I tried "hash -o
/etc/mail/access." No dice.
Nothing works and it's making me want to MURDER!
Help, please.
Here's my access file:
# Check the /usr/doc/sendmail-8.9.3/README.cf file for a description
# of the format of this file. (search for access_db in that file)
# The /usr/doc/sendmail-8.9.3/README.cf is part of the sendmail-doc
# package.
#
# by default we allow relaying from localhost...
localhost.localdomain RELAY
localhost RELAY
192.168.1 RELAY
In place of 192.168.1 I have tried individual addresses on my subnet,
192.168.1.1, etc - and I have tried covering the whole subnet,
192.168.1.0/24. No dice.
NO DICE NO DICE NO DICE!
br
@-}-
------------------------------
From: "Justin" <[EMAIL PROTECTED]>
Subject: Re: Routing problems
Date: Tue, 14 Sep 1999 14:19:15 -0600
try changeing the ip of your Beavis machine to a 10.0.0.0 address and make
sure you linux box permits access from that machine in the hosts file. also
enable ip forward and ip masq. Beavis will not be able to see the router by
itself. Linux needs to route it to it.
Matt Anderson wrote in message <[EMAIL PROTECTED]>...
>I am having some problems with routing...
>
>I have a dual-homed PC - "butthead" - with eth0(10.0.0.1) connected to the
>LAN and eth1(xxx.9.89.218) connected to the Router(210.9.89.217)
>
>"Beavis" has only one NIC - eth0(210.9.89.219) and is connected to the same
>LAN.
>
>I have added an entry to butthead's routing table to tell it that the
>router(210.9.89.217) is accessed via eth1 and an entry setting the default
>gateway(0.0.0.0) to the router.
>
>I have added to beavis' routing table an entry for default gateway(0.0.0.0)
>to point to butthead(210.9.89.218)
>
>From butthead I can ping beavis and vice versa. butthead can also ping the
>router, however beavis cannot.
>
>Any ideas on what I need to do to get the connectivity from beavis to the
>router? I'm getting desperate and have no idea what's wrong.
>
>Thanks in advance
>
>Matt
>
>
>
>
------------------------------
From: [EMAIL PROTECTED] (David C.)
Subject: Re: Why does hitting a key seem to wake up TCP/IP?
Date: 14 Sep 1999 15:30:54 -0400
Jeff Sumner <[EMAIL PROTECTED]> writes:
>
> Interestingly enough, with the included PCI Token Ring driver, the
> same effect can be seen in an x-windows session to another workstation
> except it happens much faster- if one moves the mouse on the
> x-station, the session is much faster, and tranfer rates climb very
> quickly, but if not, things take *forever* to draw. Didn't happen with
> the 16/4 T-R driver/card combo.
This is probably an application bug.
X tries to aggregate its network traffic to reduce bandwidth. When you
move the mouse, more events are generated, so the X server ends up
sending more requests to the application, which causes the application
to flush its events more quickly in response.
I say this is an application bug for several reasons:
1: Unless absolutely necessary, X apps shouldn't request every
mouse-move event from the server. It tends to bog down performance.
It is far better to just get the coordinates where click events
happen, and only track the pointer during drag operations.
2: The application's drawing code shouldn't require continuous feedback
from the server. If non-drawing code is being slowed down, then the
application is improperly using the X event loop - using it to
control non-X-related events. This can be done, but there must be a
means to unblock the event loop without the user generating X
events.
I have no idea why the choice of NIC should have any impact on this.
Bad drivers can impact performance, but it seems odd that the X display
would have any impact.
-- David
------------------------------
From: [EMAIL PROTECTED] (David C.)
Subject: Re: TCP/IP Programming Mystery
Date: 14 Sep 1999 15:44:19 -0400
[EMAIL PROTECTED] (Jerry Gardner) writes:
>
> Here's one for all of you TCP/IP sockets gurus:
>
> I've written some sockets code (in C) that implements a simple
> client-server application. The client sends 2000 byte buffers to the
> server over a TCP connection.
>
> When I run the client and server on different machines connected on an
> Ethernet (both with MTU=1500), the server's 'recv' call sometimes
> returns a byte count of 1460 bytes instead of 2000. Looking at
> Ethernet activity with snoop shows that the 2000 byte send gets
> fragmented into two IP packets: 1460 bytes and 540 bytes. The 'recv'
> call is obviously returning 1460 because that's the size of the first
> fragment. What I don't understand is why doesn't 'recv' wait until it
> gets the second fragment and return a count of 2000? (The socket is
> *not* a non-blocking socket.)
>
> When I run both the client and server on the same box, this problem
> doesn't occur, probably because the loopback device has an MTU > 2000
> bytes.
>
> I doubt this problem is specific to Linux, because I've moved the code
> to two Solaris boxes and see the exact problem.
>
> Anyone have any idea what might me happening here?
It has to do with the TCP stack on the sender side.
A properly written TCP stack will send each TCP packet at a size that
will not get fragmented. This is 1460 bytes of data:
20 bytes IP header
20 bytes TCP header
1460 bytes data
-------------------
1500 byte MTU
Theoretically, it could have sent all 2000 bytes as a single TCP packet.
In which case, the IP layer would have fragmented the packet. This
would be bad. If one of the two fragments gets lost in transmission,
both have to be retransmitted. And the receiver can't process any data
until both fragments arrive.
In general, you should expect this kind of behavior, even with smaller
size buffers. The "internet standard" is that all MTUs be 576 bytes or
larger. This means that you could send out a 1400 byte buffer, and it
would be sent at once over Ethernet, but mught be sent in 2 or 3 pieces
to a receiver over the internet if some transit network has an MTU
smaller than 1400. (For example, my ISP uses an MTU of 1009 for its
dial-up modem links.) Well written TCP stacks will discover the
end-to-end MTU and set their packet sizes appropriately.
It is important to keep in mind that TCP is a streaming protocol. You
don't send buffers. You send a stream. If you tell the sender to send
a buffer, it could be received all at once, or in any number of
various-sized pieces. (eg: send 2000, receive 1400, then receive 600).
Furthermore, buffers sent immediately after each other may be received
together. (eg: send 100, send 100, receive 200) It is always the
application's responsibility to make sure this doesn't cause ill
behavior.
If it is critical that the network transmits each buffer as a single
unit, you could use the UDP protocol. With UDP, your entire buffer (up
to 64K) is sent in a single IP datagram - which may be fragmented by the
IP stack. The receiving application will not receive anything until it
all arrives. UDP will ensure that buffers do not get broken up or
combined. The downside is that UDP is not reliable. If any fragment of
the packet gets lost in transmission, the entire packet is discarded -
you don't even get an error. It becomes the application's
responsibility to deal with the possibility of lost packets.
-- David
------------------------------
From: "Blacka" <[EMAIL PROTECTED]>
Subject: One more question
Date: Tue, 14 Sep 1999 12:25:51 -0700
> Hi, I currently have a windows 98 machine, and a pent 90 with DOS. I was
> wondering if it was possible to install linux on the DOS computer through
a
> network. Does the network have to be set up first? The hub and all the
other
> stuff are set up, but stuff like samba are not in place since linux isnt
> installed yet.
>
> My question : How do I install linux on a pentium 90 computer which has no
> modem, and no cd-rom, it only has a network card. Please help me.
>
> Orrin Jolly
>
Question : Do i have to install the network card under DOS? or can i just
install it physically and linux will recognise it?
Thank you sooo much
------------------------------
From: "Cjk Didelot" <[EMAIL PROTECTED]>
Subject: SATAN
Date: Tue, 14 Sep 1999 17:56:15 +0200
Where can I find SATAN for LINUX.
------------------------------
From: "Russ W. Knize" <[EMAIL PROTECTED]>
Subject: MSCHAP81 ???
Date: Tue, 14 Sep 1999 14:50:02 -0500
I am trying to use pptp to connect to an NT server SP5 via PPTP. I have
enabled MSCHAP80 and recompiled pppd, but when I saw the errors
initally, the NT server was asking for CHAP:81 not 80. I assume that
this is related to the RAS bugfix in SP4/5. I tried changing the chap.h
to 0x81, hoping it would work. No dice.
Any ideas?
Thanks,
Russ W. Knize
------------------------------
From: "Edward Lee Ah Yen (AKA SMASH)" <[EMAIL PROTECTED]>
Subject: Re: Good Book on Firewalling - IPChains ?
Date: Tue, 14 Sep 1999 15:49:11 GMT
http://www.ecst.csuchico.edu/~dranch/LINUX/TrinityOS.wri
http://members.home.net/ipmasq/ipmasq-HOWTO-1.77.html
daben wrote:
> Hi all
>
> Can anyone recommned some good primers on firewalling and security
> (specifically on ways of monitoring hacks). Also how about some good info
> on ipchains? I have a net up and runnig with firewalling, forwarding and
> filtering but am a little shaky on how good the security is.
>
> Thanks
> daben
------------------------------
From: "Justin" <[EMAIL PROTECTED]>
Subject: firewalling a larger network
Date: Tue, 14 Sep 1999 14:40:50 -0600
I run a larger network with multiple site. They are all connected to our
main lan via a mix of 56k frames and T1's. and are all comming in through
cisco routers from a cloud at our local phone co. Also comming from the
cloud to our main site is a 256k frame line to our isp who has also been
managing our firewall and NATfor us. We are now ready do our own
firewalling, but I am not sure as to the best way to proceed. We do have a
linux server with a static ip that our isp translates for us to send and
recieve mail. All routers are set to use our main router as their gateway of
last resort. and our main router is set to use the ip of our interface on
our isp's router as its gateway of last resort.
as i see it there are a few different options and i am not sure the best and
most secure way to proceed.
any help or direction in good examples or reading would be appreciated
Justin Workman
------------------------------
Date: Tue, 14 Sep 1999 19:31:23 +0100
From: Kevin Wall <[EMAIL PROTECTED]>
Subject: Port 106
Can anyone through any light on how a user can change their user
password using the 106 port supported by most email clients?
Is it using a POP server deamon or a telnet server deamon or something
completely different?
Any help would be gratefully received as we do not like the idea of
users having full telnet access just to change their passwords.
Regards
Kevin
------------------------------
From: [EMAIL PROTECTED] (David C.)
Crossposted-To: comp.os.linux.setup
Subject: Re: Sub-C networks?
Date: 14 Sep 1999 15:58:30 -0400
[EMAIL PROTECTED] (Andrew Brennan) writes:
>
> I have been having nothing but trouble with my Linux box and my
> local network. It's a sub-C mask (255.255.255.192) and the network
> happens to be at the high end of the range (making things simple,
> I'd expect) but it's not working. Kernel version is 2.0.34, if
> that helps.
>
> route add -net 127.0.0.0 works, I can ping 127.0.0.1 ...
>
> ifconfig eth0 x.x.189.192 netmask 255.255.255.192 broadcast
> x.x.189.255 also works ... ifconfig eth0 reflects my settings.
A useful exercise if you're using unusual netmasks is to separate out
the network and host portions of each address. In this case, you've
got:
x.x.189.192 MASK 255.255.255.192
This breaks down (in binary) into:
network: xxxxxxxx.xxxxxxxx.10111101.11******
host : ********.********.********.**000000
(where * represents bits that are part of the other portion of
the address.)
There's your problem. The host portion of your address is all zeros -
an illegal address.
Change your address from x.x.189.192 to x.x.189.193 (or any value
between 193 and 254) and all should start working again.
-- David
------------------------------
From: [EMAIL PROTECTED] (David C.)
Crossposted-To: comp.os.linux.setup
Subject: Re: Sub-C networks?
Date: 14 Sep 1999 16:01:54 -0400
Thomas Kaemer <[EMAIL PROTECTED]> writes:
>>
>> route add -net x.x.189.192 takes, I can ping my own box. I can't
>> ping anything on the local network, though.
>
> If you have a 2.2.x kernel the routing table is built automatically.
I should hope this kind of route is build automatically on all kernels.
The default route to locally-attached subnets is something every IP
stack should fill in automatically. At least I've never seen an OS
where it didn't. (including Win9x, WinNT, OS/2, SunOS, Solaris, and
various revs of Linux.)
-- David
------------------------------
From: [EMAIL PROTECTED] (David C.)
Crossposted-To: comp.sys.ibm.pc.hardware.networking
Subject: Re: Recommendation for 100Mbps Switched Ethernet hardware
Date: 14 Sep 1999 16:04:55 -0400
"Ricardo Wagner" <[EMAIL PROTECTED]> writes:
>
> Depends of
>
> - the size of the network you are running
> - The traffic
> - the number of servers
> - your budget
>
> Switches come in a variety of specs. I run a mix of HP (cheap and
> simple.. not very fast, good for small branches) and Bay Networks for
> heavy traffic... specially for our server room and large
> headquarters... flexible...expensive..but monsters.
Yep. You can get cheap 4-port unmanaged switches for around $100.
You can also get switches that can do routing, filtering and firewalling
for hundreds of ports for a price several orders of magnitude higher.
-- David
------------------------------
From: "Jerry O'Brien" <[EMAIL PROTECTED]>
Subject: Help! Inetd doesn't respond to requests.
Date: Tue, 14 Sep 1999 14:53:47 -0500
After tons of calls saying "I can't get my mail" I found it was inetd, and
not my POP3 daemon at fault. FTP is out too. If I stop and restart inetd,
everything is fine (for a few minutes). The only thing I can find in the
logs is this, which is totally greek (geek?) to me:
cleaning rbuf for sk=02141a04
sk->rspace = 14258, was 14258
repeated many times, and then occasionally this:
tcp_ack: seq 8f2b20ea ack 61e8fb7e
Data wakeup.
It seemed to start out of the blue over the weekend when no other changes
were being made.
Thanks for reading,
Jerry O'Brien
------------------------------
From: Maarten Afman <[EMAIL PROTECTED]>
Subject: Re: Problem to reach 100Mbps with 3C905B-TX
Date: Tue, 14 Sep 1999 22:10:27 +0200
Bernard Fay wrote:
> > The software should have nothing to do with this.
> >
> > Look on the back of the card. Which of the link lights is on? If the
> > 100M light is on, then you have a 100M connection. If the 10M light is
> > on, then check what's at the other end of the cable - the card will
> > auto-detect based on what it finds there. If you've got a 10/100
> > auto-sensing hub, see if you can force the port to 100M - its
> > auto-sensing algorithm may be conflicting with 3Com's, resulting in them
> > negotiating to 10M.
> >
> > If you have a 100M connection, and you're simply not seeing anything
> > close to 100Mbps of throughput, that's normal. The fastest PCs don't
> > usually manage to acheive more than 25-30Mbps on a 100M link. It's a
> > limit of the PCI bus, most Ethernet cards, and the I/O capacity of
> > current-design PC motherboards.
> >
> > (BTW, this is why a GigE card on a PC host is a waste of money. It
> > still won't get much more than 30Mbps out of the link's 1000Mbps
> > capacity. GigE is far more useful for connecting switches to each other
> > than for connecting hosts to switches.)
> >
> > -- David
Well, there is a lot wrong with the design of the PC, but it's also the ethernet
protocol itself
which is to blame for not "achieving what it says". Ethernet can be *extremely slow*
when all hosts
connected to the network all want to transmit data. Their can only be one one active
sender at a
time. The problem is that there is no mechanism to specify which host is allowed to
send. The
ethernet boards start emitting ethernet packets until a "collision" occurs i.e. they
detect that
some other device also sent packets. Then the ethernet board will wait for a short
random interval
end resume sending. This is all very unsophisticated and is inefficient for high
traffic networks.
Think of 10 Mbit networks only pouring data through at rates of 250 kB/s!
--
(( Maarten Afman ))
)) email: [EMAIL PROTECTED] ((
(( homepage: http://delft.dyndns.org ))
))
(( xxxx
------------------------------
** FOR YOUR REFERENCE **
The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:
Internet: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.networking) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Networking Digest
******************************
