Date: Mon, 9 Dec 1996 15:05:14 +1100 (EST) From: proff@suburbia.net To: bde@zeta.org.au (Bruce Evans) Cc: hackers@freebsd.org Subject: Re: strange problems with recent current Message-ID: <19961209040514.4547.qmail@suburbia.net> In-Reply-To: <199612090319.OAA12973@godzilla.zeta.org.au> from Bruce Evans at "Dec 9, 96 02:19:33 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
> >Dec 8 02:37:59 evil /kernel: cy15: 5 more silo overflows (total 3876) > >Dec 8 02:38:00 evil /kernel: cy8: 1 more silo overflow (total 7) > >... > >This has crept in sometime during the past few days. > > On what hardware? The PCI cy driver doesn't use a "fast" interrupt > handler yet, so silo overflows are quite likely for it. E.g., updating > the keyboard LEDs takes a few msec, so a few fifos full of input may > be dropped on each port whenever you hit caps lock (the fifo fills up > in 1 msec at 115200 bps). P100/Triton mother board. 4 Triton/mb IDE drives, two scsi (ISA adaptec), average 120 procs, 72Mb swap used (striped over the 4 ide drives). This configuration hasn't changed, and I don't use the console keyboard, infact..the console keyboard has been unplugged (could this have any effect?) > >Previously cy.c > >would suffer silo overflows far more rarely (once every 10-30 seconds > >on average) and overflows quantities were never over 2 or 3. > > One per day is too many. cy15 is often going at 57.6k in both directions, the others are more sporadic typical dial-up-user traffic (at 38.4). I moved cy15 onto sio1 (on board 16550A uart) Dec 9 14:58:16 suburbia /kernel: sio1: 1 more silo overflow (total 17190) Dec 9 14:58:21 suburbia /kernel: cy0: 2 more silo overflows (total 220) Dec 9 14:58:21 suburbia /kernel: sio1: 1 more silo overflow (total 17191) Dec 9 14:58:29 suburbia /kernel: sio1: 1 more silo overflow (total 17192) Dec 9 14:58:38 suburbia /kernel: sio1: 1 more silo overflow (total 17193) (uptime 16 hours) This is actually quite a bit better than the cyclades (could just be the 12 vs 15 byte byffer) but still worse than the cyclades before the "slow down". > >Response time generally has suffered, even with a very low load. Feels > >like some kind of excessive context switching delay. > > I haven't noticed any new problems here on a lightly loaded 486/33 system. > > Bruce > Load average is typically around 0.16 Cheers, Julian.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19961209040514.4547.qmail>