> On 2008-01-31 13:44, Tuc at T-B-O-H.NET wrote:
> > But would the device have sent a RST if it received a fragment
> > it couldn't route? I'm getting an actual RST from the router on the other
> > end of a WDS link towards the far end laptop.

> Some versions of Linux wrongly generate an RST in response to a bad TCP
> checksum. (Wrongly because the bad checksum is telling you you can't
> trust the TCP header and payload, so why generate an RST using values
> from the header?) I've had problems in the past with simply bad cabling
> exercising this bug.

But would it generate the RST *FORWARD* or *BACKWARD*?
So generally if .1 sends a packet through .2 which is WDS'd
to .5 for .6 ... If .1 or .2 forwarded/created a bad packet
destined for .6, would .5 deliver the RST to .1/.2 or to
..6? My situation is .1 sends, through .2 (Ethernet input,
WDS/radio output) to .5 (radio/WDS input, ethernet output)
for .6.

.5 sends the RST to .6. I've checked the checksums
I can capture off .6 for checksum issues, none were marked
that way.

I'll swap out the far side cable to see if it helps.

Thanks, Tuc
openssh-unix-dev mailing list