Date: Mon, 14 Jul 1997 13:52:07 +1000 From: Stephen McKay <syssgm@dtir.qld.gov.au> To: Stefan Esser <se@FreeBSD.ORG> Cc: syssgm@dtir.qld.gov.au Subject: Re: Diamond Fireport 40 on FreeBSD. Message-ID: <199707140352.NAA06526@ogre.dtir.qld.gov.au> In-Reply-To: <19970711202321.46313@mi.uni-koeln.de> from Stefan Esser at "Fri, 11 Jul 1997 18:23:21 %2B0000" References: <199707111710.TAA20550@monet.telebyte.nl> <19970711202321.46313@mi.uni-koeln.de>
next in thread | previous in thread | raw e-mail | index | archive | help
On Friday, 11th July 1997, Stefan Esser wrote: >On Jul 11, "W.G.T.J. Wanders" <wwanders@telebyte.nl> wrote: >> I recently installed a Diamond Fireport 40 Symbios/NCR 875 >> based UW SCSI controller in my PC at home and it has worked >> very fine since under FreeBSD 2.2.[1,2]. Only a small change >> in ncr.c was needed to get FreeBSD to see the card: >> + #define DIAMOND_FIREPORT_40_ID (0x008f1000ul) >Yes, and a patch to that effect has been applied to -current >as of June 11, but not yet been incorporated into 2.2.x (or >2.1.x). While we're on the topic... I've had one of these beasts for a couple of weeks now. Long enough to start poking about and noticing some unusual things. It has been completely stable under 2.2.2 (installing, building numerous kernels, one make world) reporting no errors to syslog. But ncrcontrol shows things that look wrong to me. Ncrcontrol reports, among other things, expected disconnects, unexpected disconnects, interrupts on the fly, and breaks. As far as I can tell, interrupts on the fly are from normally completed commands, and breaks are from all other sorts of failures. On my system, unexpected disconnects occur at exactly the same rate as breaks, and during moderate activity, there will be 1 or 2 every so often. Expected disconnects equals the interrupt rate, meaning, I suppose, that every read or write disconnects once and interrupts once. However, during an iozone test to my DCAS-34330 (4Gb UW), regular interrupts rise to about 200 while breaks and unexpected disconnects rise to 400 or 500! What is going on here? Do I really have 2 failed attempts for each success? Also, I've got an NEC 4x CD-ROM (narrow SCSI-2, negotiates 8.0Mb/s) attached. When I set the disk churning (like 'find / -print | xargs wc') it zips along doing (off the top of my head) about 100 operations/sec. When I start a 'dd if=/dev/rcd0c of=/dev/null bs=32k', the disk drops to about 20 ops/sec. There should be plenty of bus time free for these two to operate at maximum speed. The CD-ROM negotiates to 8Mb/s and is only using 600Kb/s, which is under 10%, so the other 90% should be available for the disk. The disk can only fill about 40% (8Mb/s vs negotiated rate of 20Mb/s), so that should leave my SCSI bus 50% idle. Instead, I see the the CD-ROM running at full speed, and the disk starved. What is going on? Mr Wanders, do you see this behaviour too? Confused but searching, Stephen. PS Any idea when the ncr driver will support 40Mb/s? I could help if I had any knowledge of the 875 chip. :-)
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199707140352.NAA06526>