In article <(E-Mail Removed)>,
(E-Mail Removed) says...
> In MsgID<(E-Mail Removed) m> within
> uk.comp.home-networking, 'Chris' wrote:
>
> >I've run ethereal on the connection and it would appear that the packet
> >is coming back from the laptop, albeit with a bad checksum (this is
> >normal though), but no connection is granted.
>
> Not quite sure what you mean by 'bad checksum this is normal' afaik such a
> packet will always be ignored. Something I don't understand here?
* Why am I seeing all these packets with bad checksums? How do I correct
* this?
If the capture was performed on a machine using a 3Com or Broadcom card
AND the flagged packets were sent by the capturing machine, then this
problem is most likely due to a special issue with cards configured for
TX Checksum Offload. (Note: we reproduced this issue on 3Com and
Broadcom cards, but suspect it may occur on other cards as well.)
Turning off 'TX checksum offload' in the advanced setting for the card
usually fixes this issue.
Some cards are configured by default for checksum offload. (For example,
the 3c905 and 3c920.) When this feature is enabled, the Windows TCP/IP
stack does not calculate the IP and TCP checksums but leaves them as
0x0000. (You can see this in the decoded packets.) When the packet
reaches the network card, the checksum is then calculated, inserted into
the packet, and sent out on the wire. EtherPeek gets a copy of each
outgoing packet before it is sent. EtherPeek sees the 0x0000 checksum
and believes these packets have a bad checksum, so it flags them
incorrectly.
> Given you can browse in the opposite direction it seems *some* packets are
> making it intact.
Exactly
That's what makes this so difficult to solve. I've asked draytek if
there's anything network monitoring software that is included within
their router to give me some packet analysis. However, they seem to
think snmp traps will provide all the info they need. They won't.
> Once a connection's been setup, the direction in which it was set up (who
> made the original connection to who) makes no difference to the passage of
> packets between the machines.
A stateful connection occurs, correct. And, as you've noted - i can
browse from wifi > lan perfectly. I jus't cant go the other way. I've
tried many differing clients and OS's (debian, gentoo and windows) and
the problem is completely replicable... :/
> >telnet <wifi laptop> 445
> >just times out. Whereas if I plug in an ethernet cable and do 'telnet
> ><ether laptop> 445' i get connected!
> >
> >Just what is going on? 
>
> Can you ping your wifi machine from the others and can you ping the others
> from the wifi?
Absolutely.
> And, have you tried running any other server program on the
> laptop?
Indeed I have. FTP, NNTP, SMTP etc all works fine when connecting in to
my laptop. I am at a loss to know why 139/445 doesn't work

It would appear to be a local firewall issue, but i have tried another
wifi client and the same thing happens. So, it /must/ be a draytek
firewall issue?
> I assume there isn't any firewalling distinction between the two
> interfaces (normal ethernet/wifi)? That'd be the one of the first things
> I'd look for if it was in front of me. No difference in IP/subnet settings
> between the two adaptors?
Absolutely none. The IP is different, but it's on the same subnet and
is in the same class range.
> Doubtless someone more knowledgable will be along shortly, but in the
> meantime it's got me intrigued if you're still about.
I've got in touch with Draytek about this, and will report back anything
thing else that might assist.
Cheers for your reply,
Chris