Date: Wed, 3 May 2006 23:59:09 -0700 From: "David G. Lawrence" <dg@dglawrence.com> To: Scott Long <scottl@samsco.org> Cc: protagonist@gmx.net, marc@enginet.com, freebsd-amd64@freebsd.org Subject: Re: nve(4) timeout fix Message-ID: <20060504065909.GC12815@tnn.dglawrence.com> In-Reply-To: <4452A98E.8020501@samsco.org> References: <20060427120035.D7FA016A43B@hub.freebsd.org> <4450D0E8.4000403@uchicago.edu> <vita-brevis-breviter-in-brevi-finietur-mors-venit-velociter-quae-neminem-veretur-868xppk0tz.fsf@shodan.gothgoose.net> <200604281646.46451.jhb@freebsd.org> <44528643.4050103@rogers.com> <4452A98E.8020501@samsco.org>
next in thread | previous in thread | raw e-mail | index | archive | help
> >Why is it too late? If it fixes a known stability problem, then why cant > >it be commited? If no commits are allowed during the -RC phase, then > >what is 6.1-RELEASE waiting for? > > > > Because there is always a last minute stream of bug fixes, and when you > wait for one, another one comes up that you wait for, then another, etc, > etc. At some point we have to draw the line. Also, it's good to let > changes shake out for a few days before cutting a release. It's not > uncommon for serious problems to be exposed several days after a commit. > SO again, we have to draw the line somewhere. It appears that the fix in -current isn't quite right in any case. A friend of mine found that transmit interrupts don't appear to be occuring at all. The only reason that the driver appears to work is that receive packets/interrupts usually occur often enough so that the transmit run-down still occurs often enough...except when it doesn't. If you do a simple test of (only) sending UDP datagrams, for example, the driver will eventually run out of buffers (or TX descriptors? - not sure which), and the driver will be wedged up with no buffer space until a packet is received (e.g. by a ping from another host). With the patch in current, this reduces the no buffer space dead time to eight seconds (the watchdog timeout), but that is still quite broken. I don't have the documentation for the nve programming API, otherwise I'd be tempted to try to figure out why it is broken. Maybe someone who has access to that info can look into this? -DG David G. Lawrence President Download Technologies, Inc. - http://www.downloadtech.com - (866) 399 8500 The FreeBSD Project - http://www.freebsd.org Pave the road of life with opportunities.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060504065909.GC12815>