Network cable disconnect

[Camilo Alejandro Arboleda]

On 6/20/07, Eric Norum <norume at aps.anl.gov> wrote:
>
> The original intent of the Internet was to be able to get data
> through even in the event of a nuclear attack.  Allowing a minute for
> a damaged connection to be cleared or routed around is quite
> reasonable under those circumstances.   What is your need to detect a
> disconnected cable that quickly?


Because the packets send before the drivers detect disconnection, appear to
be properly transmitted at application layer, but they weren't.

I'm going to enable network driver debug, and see what happens.

On Jun 20, 2007, at 11:26 AM, Joel Sherrill wrote:
>
> > Camilo Alejandro Arboleda wrote:
> >> Hello,
> >>
> >> We working with a PC386 based board, with NE2000 compatible network
> >> adapter. I've seen that disconnecting the network cable doesn't cause
> >> the "write()" function to fail immediately, but it takes about a
> >> minute to detect the network disconnection. Is it possible to detect
> >> cable faults faster? Am I doing something wrong?
> >
> > Did you turn on debug in the driver and try it?
> >
> > Is there any output at all from the driver now?
> >
> > I don't see anything explicitly in the driver about cable disconnect.
> > I do see ne_watchdog which is hooked to if_watchdog.  The
> > associated if_timer value is 2 which I think is 2 seconds based
> > upon IFNET_SLOWHZ and hz being rtems_bsdnet_ticks_per_second.
> > But this looks like a Tx timeout and is way shorter than what you
> > are seeing.
> >
>
> --
> Eric Norum <norume at aps.anl.gov>
> Advanced Photon Source
> Argonne National Laboratory
> (630) 252-4793
>
>
>


-- 
Algo mas sobre Camilo Alejandro?
http://camiloaa.blogspot.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/users/attachments/20070620/e680e9cd/attachment.html>