On Sun, Dec 11, 2005 at 06:53:35PM -0800, Stephen Hemminger wrote: > David Woodhouse wrote: > > >On Sun, 2005-12-11 at 12:02 -0800, Nathan Grennan wrote: > > > > > >> I do see that the driver can be loaded without a drop in performance, > >>it is only activating the driver that triggers it. Another thing I see > >>is that deactivating it doesn't fix the problem. In that I can go from > >>run level 5 to run level 1, where networking is shut down in the > >>process, but the performance in run level 1 will still be slow. > >> > >> Another potential clue maybe that I didn't see such a drop with raid0 > >>performance when I started using the sky2 driver. So I suspect it is > >>just an interaction between the raid5 code and the sky2 code. > >> > >> > > > >Just before leaving home last week I noticed that sky2 was causing an > >interrupt storm on my new toy, which explained the massive performance > >problems I saw there. Since I'd just taken the sky2.[ch] from the netdev > >git tree and compiled them against the Fedora kernel I didn't report it; > >just intended to look into it further when I got home. > > > > > > > Doesn't anybody read the mailing lists... There was a missing interrupt > clear that was fixed about 1 day later. That patch didn't make it into FC4.netdev.4. I'm spinning FC4.netdev.5 now, w/ just that patch added. John -- John W. Linville linville@xxxxxxxxxx -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
