strange NAT behaviour
Divacky Roman
xdivac02 at stud.fit.vutbr.cz
Thu Jul 1 08:42:03 PDT 2004
On Thu, Jul 01, 2004 at 05:22:34PM +0200, Max Laier wrote:
> On Thursday 01 July 2004 16:32, Divacky Roman wrote:
> > I have booted -O universe and it works... so there is definitely some
> > optimalization error - is it worth it to catch the bug? I'd say YES!
>
> I agree.
>
> > (isnt here even plan for officially supporting -O2?)
>
> Well, you should provided some more specific explanation how you use this box.
> i.e. how do you do NATting (ifpw + natd, ipf, pf or pppd's builtin)? What
> interfaces do you employ? What is the MTU of your outgoing path? Moreover
> tcpdumps are of interest in order to see why things do fail. Same for
> statistics (netstat -ssp {ip, tcp, udp ...}) to learn why the packets are
> being dropped.
ipfw + natd,
natd on vr0 with mtu 1500, iface to internal network is xl0 with mtu
1500
I cannot provide mtu since I have to use that machine (so I wont switch
back to -Os)
from the nat machine I was able to do all networking so I wouldnt blame
any driver, from machine behind nat ping -s 1472 worked so I wouldnt
also blame ipfw - it really seemd like fragmenting problem somewhere in
natd...
btw: ipfw shown me that some packets go to divert socket (where natd is
binded) and some to allow rule so I suppose ipfw is working
> > thnx for attention
> >
> > roman
> >
> > On Thu, Jul 01, 2004 at 03:50:25PM +0200, Divacky Roman wrote:
> > > I have upgraded from CFLAGS=-O ARCH=p2 may 17th current (both kernel
> > > and userland) to CFLAGS=-Os (I know its NOT supported) ARCH=athlon-xp
> > > 30th June current and strange thing happend
> > >
> > > The machine in charge is working as a NAT and machines behind that are
> > > not able to transfer any packet longer than 1472 bytes. its seems
> > > fragmenting is broken but on that nat machine I am able to transfer
> > > anything...
> > >
> > > I dont ask for help (since -Os is unsupported, but I am compiling -O
> > > universe now so I might ask for help ;) ) just point it as interesting
> > > behaviour... maybe some kind of bug?
> > > and YES I have it all set properly
> > >
> > > roman
>
> --
> Best regards, | mlaier at freebsd.org
> Max Laier | ICQ #67774661
> http://pf4freebsd.love2party.net/ | mlaier at EFnet
More information about the freebsd-current
mailing list