Re: ndis(4) patch to replace obsolete if_watchdog interface

On Thursday 29 May 2008 05:46:05 pm Andrew Thompson wrote:
> On Thu, May 29, 2008 at 04:41:32PM -0400, Coleman Kane wrote:
> > Hi,
> > 
> > I just replaced the obsoleted if_watchdog interface in ndis(4) with a
> > local implementation. This should remove the obnoxious warning message
> > on device init. Anyone using -CURRENT with an ndis card, could you send
> > me success/fails?
> > 
> > The patch is here:
> >   * http://people.freebsd.org/~cokane/patches/if_ndis-new_wd.patch
> 
> 
> This works different to the rest of the network drivers. The existing
> drivers use a callout tick that runs while the driver is up and an
> integer counter.
> 
>  if (x && --x == 0)
> 	 ...timeout...
> 
> You arm the callout and stop it after each Tx, does this have any
> perfornace impact?

Hmm, I'm not sure, but I'm doing the simple method myself in the various 
drivers I'm currently updating.  I would err on the side of simplicity unless 
it is proven that a more complex solution (i.e. having a separate timer that 
gets maintained) is needed.  Another thing to consider is that you now have 
the overhead of running the callout periodically in the complex case whereas 
in normal operation it would always be cancelled and never run in the simple 
case (timeouts are rare).

Probably best is to do some actual measurements both ways though.  In general, 
simpler, more intuitive alogrithms are better than more complex ones unless 
there is a real performance benefit from the complexity.

-- 
John Baldwin