Re: network slowness/freez-up since update 10/11

On Thu, 14 Oct 2004, Andrey Chernov wrote:

> On Thu, Oct 14, 2004 at 02:59:34PM +0200, Ian FREISLICH wrote:
> > but zero response IIRC.  Maybe making it on by default was a little
> > hasty, but anyone that follows -CURRENT like they should if they
> > run it weuld have been aware of this and set debug.mpsafenet=0 in
> > their loader.conf when they saw that commit.
> 
> There is nowhere said in this commit that if_de would or ever may be
> broken. Everydays something new coming, and this commit is one between
> similar others, nothing brings special attention to be aware of. I'll
> expect special 'WARNING: de owners' message to current in such
> situations, i.e. if something is not fixed to work together. 

The commit was not known in advance to break if_de, as it was believed
that the IFF_NEEDSGIANT compatibility code would cover it, as it appears
to do successfully for other drivers that are not yet safe without the
Giant lock.  While there had been sporadic reports of problems with if_de
on an SMP alpha and one other system, it was only in the last couple of
weeks that I've seen a larger number of reports.  Obviously, we're working
to fix it, but thus far it hasn't proven to be any of the more obvious
potential issues, and is likely a race in the driver that may require some
more deep reading and instrumentation to track down.

I don't recall seeing a bug report until I saw your "me to" in the last
day or two, or I probably would have been e-mailing you asking to try
various things to see if they improve behavior.  However, as I've
mentioned, I've moved continents over three weeks, so I could well have
missed your bug report.

Robert N M Watson             FreeBSD Core Team, TrustedBSD Projects
robert_at_fledge.watson.org      Principal Research Scientist, McAfee Research