Re: Rsync access through NAT?

Re: Rsync access through NAT?

[Michel Lanners]

Hi again,

To the list: for some time now, I've had problems accessing
penguinppc.org from my PPC machines (I've noticed it with rsync... for
www I use a proxy on i386).

Turns out the problem only occurs with 2.4 kernels:

[mlan@piglet ~]$ uname -a
Linux piglet 2.4.3-pre3 #4 Thu Apr 12 01:47:59 CEST 2001 ppc unknown
(benh's tree)

and 2.4.4-pre1 from benh on the TiBook.

Anyone else seen this? Note, with a 2.2 kernel, all is OK. I've not had
a chance to test a 2.4 kernel on i386 to see whether it's the same over
there.

I hope it doesn't matter, but I'm accessing the net through a NAT router
(which is configured OK, thanks for asking ;-)

Michel

-------------------------------------------------------------------------
Michel Lanners                 |  " Read Philosophy.  Study Art.
23, Rue Paul Henkes            |    Ask Questions.  Make Mistakes.
L-1710 Luxembourg              |
email   mlan@cpu.lu            |
http://www.cpu.lu/~mlan        |                     Learn Always. "


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Rsync access through NAT?

[jeramy b smith]

We had problems serving rsyncs with 2.4 also. Using 2.4 and an rsync built
against 2.2 headers, we would get hung rsync processes that a stack trace
revealed were hung in a newselect (Paulus already knows this bit).

If rsync was built against 2.4 headers it would hang on the first hung
process and then refuse connections.

-jeramy b smith




On Tue, 17 Apr 2001, Michel Lanners wrote:

> Hi again,
>
> To the list: for some time now, I've had problems accessing
> penguinppc.org from my PPC machines (I've noticed it with rsync... for
> www I use a proxy on i386).
>
> Turns out the problem only occurs with 2.4 kernels:
>
> [mlan@piglet ~]$ uname -a
> Linux piglet 2.4.3-pre3 #4 Thu Apr 12 01:47:59 CEST 2001 ppc unknown
> (benh's tree)
>
> and 2.4.4-pre1 from benh on the TiBook.
>
> Anyone else seen this? Note, with a 2.2 kernel, all is OK. I've not had
> a chance to test a 2.4 kernel on i386 to see whether it's the same over
> there.
>
> I hope it doesn't matter, but I'm accessing the net through a NAT router
> (which is configured OK, thanks for asking ;-)
>
> Michel
>
> -------------------------------------------------------------------------
> Michel Lanners                 |  " Read Philosophy.  Study Art.
> 23, Rue Paul Henkes            |    Ask Questions.  Make Mistakes.
> L-1710 Luxembourg              |
> email   mlan@cpu.lu            |
> http://www.cpu.lu/~mlan        |                     Learn Always. "
>


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Rsync access through NAT?

[Michel Lanners]

Hi Jeramy,

On  17 Apr, this message from jeramy b smith echoed through cyberspace:
>
> We had problems serving rsyncs with 2.4 also. Using 2.4 and an rsync built
> against 2.2 headers, we would get hung rsync processes that a stack trace
> revealed were hung in a newselect (Paulus already knows this bit).

Might this be the problem here:

http://slashdot.org/articles/01/04/24/0255224.shtml

I have ECN compiled in all my 2.4 kernels. They are building right now
without ECN.... We'll see.

<a compile & reboot later>

Yup, that seems to be the problem. rsync works again for me on
penguinppc. So I guess that some device close to penguinppc.org doesn't
support ECN.....

Well, no biggy, I'll just leave it disabled for now...

Cheers

Michel

-------------------------------------------------------------------------
Michel Lanners                 |  " Read Philosophy.  Study Art.
23, Rue Paul Henkes            |    Ask Questions.  Make Mistakes.
L-1710 Luxembourg              |
email   mlan@cpu.lu            |
http://www.cpu.lu/~mlan        |                     Learn Always. "


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Rsync access through NAT?

[Ethan Benson]

[-- Attachment #1: Type: text/plain, Size: 698 bytes --]

On Tue, Apr 24, 2001 at 07:53:38AM +0200, Michel Lanners wrote:

> I have ECN compiled in all my 2.4 kernels. They are building right now
> without ECN.... We'll see.
> 
> <a compile & reboot later>

there is a sysctl variable that disables this at runtime you know ;-)
i don't remember where it is offhand, sysctl -a | grep ecn or find
/proc -name '*ecn*' -print should find it.  

> Yup, that seems to be the problem. rsync works again for me on
> penguinppc. So I guess that some device close to penguinppc.org doesn't
> support ECN.....

probably turning on ecn is more trouble then its worth right now,
unfortuanatly... 

-- 
Ethan Benson
http://www.alaska.net/~erbenson/

[-- Attachment #2: Type: application/pgp-signature, Size: 0 bytes --]

Re: Rsync access through NAT?

[Geert Uytterhoeven]

On Mon, 23 Apr 2001, Ethan Benson wrote:
> On Tue, Apr 24, 2001 at 07:53:38AM +0200, Michel Lanners wrote:
> > I have ECN compiled in all my 2.4 kernels. They are building right now
> > without ECN.... We'll see.
> >
> > <a compile & reboot later>
>
> there is a sysctl variable that disables this at runtime you know ;-)
> i don't remember where it is offhand, sysctl -a | grep ecn or find
> /proc -name '*ecn*' -print should find it.
>
> > Yup, that seems to be the problem. rsync works again for me on
> > penguinppc. So I guess that some device close to penguinppc.org doesn't
> > support ECN.....
>
> probably turning on ecn is more trouble then its worth right now,
> unfortuanatly...

| tux$ cat /proc/sys/net/ipv4/tcp_ecn
| 1
| tux$

I have it enabled on all my machines running 2.4.x. So far no problems.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Rsync access through NAT?

[Michel Dänzer]

Geert Uytterhoeven wrote:
>
> On Mon, 23 Apr 2001, Ethan Benson wrote:
> > On Tue, Apr 24, 2001 at 07:53:38AM +0200, Michel Lanners wrote:
> > > I have ECN compiled in all my 2.4 kernels. They are building right now
> > > without ECN.... We'll see.
> > >
> > > <a compile & reboot later>
> >
> > there is a sysctl variable that disables this at runtime you know ;-)
> > i don't remember where it is offhand, sysctl -a | grep ecn or find
> > /proc -name '*ecn*' -print should find it.
> >
> > > Yup, that seems to be the problem. rsync works again for me on
> > > penguinppc. So I guess that some device close to penguinppc.org doesn't
> > > support ECN.....
> >
> > probably turning on ecn is more trouble then its worth right now,
> > unfortuanatly...
>
> | tux$ cat /proc/sys/net/ipv4/tcp_ecn
> | 1
> | tux$
>
> I have it enabled on all my machines running 2.4.x. So far no problems.

You're lucky, it used to keep me from establishing any TCP connections to the
outside world through an ISDN router.


--
Earthling Michel Dänzer (MrCooper)    \   Debian GNU/Linux (powerpc) developer
CS student, Free Software enthusiast   \        XFree86 and DRI project member

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Rsync access through NAT?

[jeramy b smith]

I'm not sure. The routers are cisco 55xx and 75xx with the latest firmware loaded. From there it hits the SprintLink and Abovenet clouds. I wish there was a tool that could do an point to point traceroute like journey and stop when it hits a hop that doesn't support ECN. Or does traceroute do this when used with 2.4 and ECN support on?

-jeramy b smith

mlan@cpu.lu wrote:
>
>
> Yup, that seems to be the problem. rsync works again for me on
> penguinppc. So I guess that some device close to penguinppc.org doesn't
> support ECN.....
>
--
-jeramy b smith


"Linux is Fun!" -Linus Torvalds


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Rsync access through NAT?

[Geert Uytterhoeven]

On Tue, 24 Apr 2001, jeramy b smith wrote:
> I'm not sure. The routers are cisco 55xx and 75xx with the latest firmware loaded. From there it hits the SprintLink and Abovenet clouds. I wish there was a tool that could do an point to point traceroute like journey and stop when it hits a hop that doesn't support ECN. Or does traceroute do this when used with 2.4 and ECN support on?

I think it does.

> -jeramy b smith
>
> mlan@cpu.lu wrote:
> > Yup, that seems to be the problem. rsync works again for me on
> > penguinppc. So I guess that some device close to penguinppc.org doesn't
> > support ECN.....

Then why can I ping penguinppc.org without problems? Because my AXP firewall
still runs 2.2.x?

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Rsync access through NAT?

[Michel Lanners]

Heya,

On  24 Apr, this message from Geert Uytterhoeven echoed through cyberspace:
> On Tue, 24 Apr 2001, jeramy b smith wrote:
>> I'm not sure. The routers are cisco 55xx and 75xx with the latest firmware loaded.

Fully Cisco on my end also... 803 as a NAT access-router at home, and
36xx devices from there on.

> From there it hits the SprintLink and Abovenet clouds. I wish there
was a tool that could do an point to point traceroute like journey and stop when it hits a hop that doesn't support ECN. Or does traceroute do this when used with 2.4 and ECN support on?
>
> I think it does.

No idea here... I'm new to ECN; I did some lookups at work today but
didn't have time to finish reading on it....

>> mlan@cpu.lu wrote:
>> > Yup, that seems to be the problem. rsync works again for me on
>> > penguinppc. So I guess that some device close to penguinppc.org doesn't
>> > support ECN.....
>
> Then why can I ping penguinppc.org without problems? Because my AXP firewall
> still runs 2.2.x?

Maybe because the NAT function on my Cisco 803 is to blame? Cisco might
'accidentaly' support it in normal packet forwarding, but might not when
doing NAT, since the IP header may need to be rewritten-- probably
without the ECN bits.

Cheers

Michel

-------------------------------------------------------------------------
Michel Lanners                 |  " Read Philosophy.  Study Art.
23, Rue Paul Henkes            |    Ask Questions.  Make Mistakes.
L-1710 Luxembourg              |
email   mlan@cpu.lu            |
http://www.cpu.lu/~mlan        |                     Learn Always. "


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Rsync access through NAT?

[Michel Dänzer]

Michel Lanners wrote:

> >> > Yup, that seems to be the problem. rsync works again for me on
> >> > penguinppc. So I guess that some device close to penguinppc.org doesn't
> >> > support ECN.....
> >
> > Then why can I ping penguinppc.org without problems? Because my AXP
> > firewall still runs 2.2.x?
>
> Maybe because the NAT function on my Cisco 803 is to blame?

NAT is also a suspect of mine - it's the commonality between our cases.

> Cisco might 'accidentaly' support it in normal packet forwarding, but might
> not when doing NAT, since the IP header may need to be rewritten-- probably
> without the ECN bits.

I must admit I don't really know anything about ECN but that the name of the
/proc file is tcp_ecn, right? It kinda implies that ECN is only involved with
TCP, not ICMP or UDP or whatever. In particular, UDP worked fine (DNS lookups)
when TCP didn't.


--
Earthling Michel Dänzer (MrCooper)    \   Debian GNU/Linux (powerpc) developer
CS student, Free Software enthusiast   \        XFree86 and DRI project member

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Rsync access through NAT?

[Peter Cordes]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=unknown-8bit, Size: 1877 bytes --]


On Tue, Apr 24, 2001 at 11:59:17PM -0500, Linux PPC Developer Digest wrote:

> Date: Tue, 24 Apr 2001 22:51:10 +0200
> From: Michel =?iso-8859-1?Q?D=E4nzer?= <michdaen@iiic.ethz.ch>
> Subject: Re: Rsync access through NAT?

> > Cisco might 'accidentaly' support it in normal packet forwarding, but might
> > not when doing NAT, since the IP header may need to be rewritten-- probably
> > without the ECN bits.
>
> I must admit I don't really know anything about ECN but that the name of the
> /proc file is tcp_ecn, right? It kinda implies that ECN is only involved with
> TCP, not ICMP or UDP or whatever. In particular, UDP worked fine (DNS lookups)
> when TCP didn't.
>
> - --
> Earthling Michel Dänzer (MrCooper)    \   Debian GNU/Linux (powerpc) developer
> CS student, Free Software enthusiast   \        XFree86 and DRI project member

 That's correct.  The ECN flags go in the TOS byte of the IP header.  (octet
in network-speak).  The rationale is that a fixed location in the IP header
is faster for routers to deal with than a TCP option.  (Also, ECN could
obviously be used for any protocol that runs over IP, not just TCP.  AFAIK,
people currently are only using it for IP packets that hold TCP segments.)

 See http://www.faqs.org/rfcs/rfc2481.html for info on ECN in IPv4.  Search
on "ecn" at http://www.faqs.org/rfcs/ to turn up some other more recent
RFCs.  I haven't read them yet.  (BTW, especially read section 19 of rfc
2481, it explains the history of different definitions of the TOS byte.)

--
#define X(x,y) x##y
Peter Cordes ;  e-mail: X(peter@llama.nslug. , ns.ca)

"The gods confound the man who first found out how to distinguish the hours!
 Confound him, too, who in this place set up a sundial, to cut and hack
 my day so wretchedly into small pieces!" -- Plautus, 200 BCE

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/